AU Class
AU Class
class - AU

Knowing the Future Today: Predicting Safety Risk with AI and BIM 360

このクラスを共有
ビデオ、プレゼンテーション スライド、配布資料のキーワードを検索する:

説明

There is a well-established precedent for predictive analytics in business. In 2009, Google built a predictive model using years of data to determine which employees were most likely to quit. Suffolk and Autodesk partner Smartvid.io partnered to explore if predictive analytics could be applied to safety in construction. In this class, Jit Kee Chin, Suffolk’s EVP and chief data officer, and Josh Kanner, founder and CEO of Smartvid.io, will discuss how 10 years of BIM 360 Field software and other data sources were combined to create a safety predictive model. We’ll show examples of the data, discuss machine-learning frameworks for computer vision as well as predictive analytics, and we’ll present the results of the modeling exercise, how it can be displayed to the user, and what may come next. Attendees will participate in a discussion on what this new ability to see into the future could mean for the operations, human resources, and legal structures of construction?

主な学習内容

  • Understand the difference between machine learning for observation (for example, computer vision) and prediction
  • Learn how to connect your BIM 360 data to external systems to extract additional safety insights
  • Learn past examples and gain ideas for the future of predictive analytics in practice
  • Learn how to impress colleagues, friends, and family with an example of AI for construction

スピーカー

  • Joshua Kanner
    Josh Kanner is Founder & CEO of Smartvid.io, an AI-powered cloud platform that helps companies identify highest risk projects and act to prevent incidents from occurring.. Most recently he was co-founder of Vela Systems, a pioneer in the use of web and tablet workflows for construction and capital projects. There he led the company’s product, marketing and business development functions. Vela Systems grew from bootstrapped beginnings to include over 50% of the ENR Top Contractors as customers and deployments all over the globe. The company was successfully acquired by Autodesk in 2012 and has been rebranded as BIM 360 Field. Prior to founding Vela Systems, Josh was responsible for product management and strategy at Emptoris (now part of IBM), a web-based strategic sourcing software company with customers including Motorola, GlaxoSmithKline, Bank of America, and American Express. He still gets excited to put on a hard hat and walk a job.
  • Jit Kee Chin
    Jit Kee Chin is the Chief Data Officer and Executive Vice President at Suffolk and is responsible for leveraging big data and advanced analytics to improve the organization’s core business. Ms. Chin is also responsible for helping to position Suffolk to achieve its vision of fundamentally reinventing the future of construction in the digital age, working closely with the company’s Innovation and Strategy teams. Prior to her role at Suffolk, Ms. Chin spent 10 years with management consulting firm McKinsey and Company where she counseled senior executives on strategic and commercial topics. Most recently, she was a Senior Expert in Analytics in McKinsey’s Boston office where she specialized in the design and implementation of end-to-end analytics transformations. Prior to that role, Ms. Chin was an Associate Principal in McKinsey’s London office. Ms. Chin holds a PhD in Physics from the Massachusetts Institute of Technology and a BS from the California Institute of Technology.
Video Player is loading.
Current Time 0:00
Duration 55:07
Loaded: 0.30%
Stream Type LIVE
Remaining Time 55:07
 
1x
  • Chapters
  • descriptions off, selected
  • en (Main), selected
Transcript

NICHOLAS CARBONE: I'm Nicholas Carbone, and this is "Knowing the Future Today: Predicting Safety Risk with AI and BIM 360."

JOSH KANNER: I'm Josh Kanner I'm with Smartvid.IO We partnered with Suffolk on the results you're going to see. We're going to present the real results of a collaboration that we've been doing, which goes back many, many years. But this specific work that we're going to be presenting on actually goes back over a year that we've been working on this stuff together. And this will be the most detailed presentation of the results of this work together that we've given in any venue. So we're really excited about it.

Just a quick show of hands, folks in the audience, how many folks went to the AEC keynote today? Most folks? OK. So if you saw, I think his name was Andy Leek from PARIC. You guys saw him talk about predictive analytics and how it complements what's going on in the BIM 360 project dashboard.

This is very much in that vein. Whereas what BIM 360 is doing is looking at text for predictive analytics, we're complementing that by looking at what's in the photos and trying to help drive predictions off of what's in that data. So if you're wondering how the two fit together, that's how. And if you've gotten a preview from the BIM 360 insights team while you've been here at AU, you may have seen a little mock up of the predictive results that we're showing you in their user interface, which is an exciting thing that's being worked on. So with that as context, I'm going to turn it over to Nick. He'll be your tour guide for today.

NICHOLAS CARBONE: In the beginning.

JOSH KANNER: And we'll get right into the material.

NICHOLAS CARBONE: So, yeah. Thank you all for being here. And what we're going to do, as Josh said, is we're going to talk about how we combined 10 years of BIM 360 field data, as well as some additional information, in order to create a truly predictive safety model for our projects. We're going to show some examples. We're going to discuss the machine learning frameworks that Josh and his team and put together. And then we're going to present the results and what's coming next with that.

So the real practical application of this session is that you're going to walk out of here knowing that you're going to be able to have a safer job site and safer trade partners, not in some sort of nebulous future, but very, very soon. So, as I said before, I'm Nicholas Carbone. I'm a data scientist at Suffolk. I worked very closely with Josh and his team on this project from the Suffolk side. And Josh, do you want to introduce yourself a little bit more detail?

JOSH KANNER: Sure. Sure. I'm Josh Kanner. Who here in the audience has used BIM 360 Field? Hopefully a lot of folks. Yeah. So I was the co-founder of a company called Vela Systems that was acquired and became BIM 360 Field. So I've been doing construction technology since 2005, primarily in the field, going out and looking at how. Back in 2005 we were bringing mobility in the cloud to construction and seeing how that could be used to transform field processes and construction management processes like quality, safety, commissioning, and more.

And now here at Smartvid.IO what we're doing is, instead of mobility in the Cloud, which were the new technology enablers back in 2005, now in 2018 some of the really exciting technology enablers are artificial intelligence, machine learning. Because now we're in this really amazing time where because of systems like BIM 360 Field we have more data than ever to actually analyze. And the question is, what can you do with it?

And we're going to talk you through the journey that we and Suffolk went on to move from using that data to better observe what's going on today, to you using it to predict what's going to happen in the future. Just one other note, you were originally slated to have Chris Mayor as your speaker from Suffolk, but you guys lucked out because Nick actually did the work. So you're going to get a much like deeper level of detail here in terms of what we did together. So it's lucky it worked out that way.

NICHOLAS CARBONE: Yeah. So for those of you who may not be familiar with us at Suffolk, we're about a three billion plus national GC with headquarters in Boston. We have projects and offices in Boston, New York, Miami, Tampa, Dallas, San Diego, Los Angeles, and San Francisco.

And now as a GC, we're in the business of making each project successful. And I'm not sure if anyone-- anyone out there from GCs? Great. And we are all in the project-- in the business of making project successful. And to do so as GCs, we need to manage risk.

And there are five main areas of risk that Suffolk and every GC have to manage in order to stay both competitive, and in order to deliver exceptional value to our customers. Those are safety, schedule, cost, quality, and experience. And to deliver truly a transformative GC experience we need to make sure that every single partner, our owners, subcontractors, architects, engineers, project managers, field team, and trade partners subcontractors, they all have a great experience with Suffolk. And when we properly manage that risk, we succeed in our goals, which is to build fast, to build quality, and to build right. Basically to build smart. And that's a very simple concept that's very difficult to achieve. And that is where truly managing the risk becomes incredibly important.

And to truly manage risk we also need to truly understand our projects. And understanding projects requires us to understand and have our data. But not just any data. Data by itself is not smart.

Smart data requires process, technology, and people working together in a way that can be leveraged to manage all that risk. So how are we doing that at Suffolk? We've invested in innovation infrastructure in all seven of our key regional offices.

These smart labs are where it all happens. We bring together the process, the technology, and the people in conjunction with our customers, employees, and project teams to develop new technologies and new solutions. We're also able to leverage the smart labs within the local innovation landscape in order to conduct experiments with new products and technologies on our job sites.

Now all that technology can be really cool, but it's not very useful in a vacuum. And without active engagement from everyone in the organization, from the front office to the back office, project personnel to field teams, to the executive teams, as well as all of our partners, we won't be able to drive change across the organization.

Now the key is to translate these pilots and these projects from our regional smart labs when they're successful into our national toolkit, and then spread it out across the entirety of our national portfolio. This means that our tools can't be static. Just as we ourselves have to drive ourselves to be continuously improving, we have to drive our tools and our processes that way.

So after we've innovated and piloted our original smart labs we take our successful projects and roll them out into our national best practices. And I could go into detail on that whole process, but fundamentally it begins with something to test. And that's where having smart innovative partners is really, really important.

As Josh already talked a little bit about, SmartGood has been one of those partners. We first met Josh back through Vela Systems in 2005. We've been piloting their Smartvid since 2015, using their field observation technology as well as their safety monitoring technology.

But about a year ago, we came together and we sort of challenged each other to go beyond that. We said, we have such great swath of information. You're doing great things with the images. We've got all sorts of detailed information in-house. What can we do to drive this from observational to truly predictive? And so that's what we're here to talk about today, what we've done, and what's going on in the future-- with it in the near future.

It's garnered some attention in the press, et cetera, et cetera. But right now we're going to talk a little bit more detail about it with all of you. So Josh, do you want to talk a bit about Smartvid?

JOSH KANNER: Sure. Thanks, Nick. So I'm going to give a little bit of the history of Smartvid, and then we'll get into the results of the work we did together on predictive analytics. By the way, the press that Nick mentioned as yada yada, it's actually one of the more exciting moments of my professional career as a Boston guy.

NICHOLAS CARBONE: I'm so sorry.

JOSH KANNER: You didn't go to MIT, right?

NICHOLAS CARBONE: No.

JOSH KANNER: No. So, yeah. So we were featured in an article at MIT Technology, which I've always wanted to be in MIT tech review. I'm an alum. And so we were in there. And it was pretty-- it was pretty awesome. It was Nick and his boss, Jit Kee Chin, who's the executive vise president and chief data officer for Suffolk, which is interesting. They're one of the few firms I've seen that actually has someone in the executive management team who has the title of chief data officer. So, anyway, if you're interested, we can-- actually, if you just Google MIT tech review Suffolk the article will come right up.

NICHOLAS CARBONE: I apologize for dismissing you're greatest professional moment.

JOSH KANNER: Nope. That's fine. I'm used to it. As I said, I've been married for 21 years. So-- So, great. So Smartvid, real quick. What are we doing? We're trying to make AI, this new enabling technology, as easy to use as possible in the AEC industry to attack some of those drivers of risk that Nick talked about, safety, productivity, and quality.

So AI is everywhere. You hear about it all the time. I like showing this quote, because I think it shows a couple of things. One, how crazy people are getting about describing the potential of AI. But two, how respected those people who are seemingly being crazy about it.

So this is Sundar from-- you know, the Google CEO. Sundar Pichaie from Google. He says, "AI is one of the most important things humanity is working on. It's more profound than electricity or fire." And the dot, dot, dot, by the way, it's not-- there's no modifiers in there. It's just, I shortened it so you could get to electricity or fire.

Clearly, this is a little bit overstated. I'd much rather be in a house that had electricity and fire and no AI, than in a house that was dark and cold and had AI. But you know, maybe that's like Maslow's hierarchy of needs or something like that. They put-- he puts AI bottom.

But it is-- there is actually some substance to what he's saying. So if you read about what Google has been doing over the course of the last five years, they've been systematically moving from statistical based methods for doing things like even their core search engine, and replacing them with deep learning based models, AI based models. So the search engine that, you know, Serjay and Larry made for Google and built Google off of is no longer in use to do search engine optimization there-- or search engine results. It's now an AI based or deep learning based model. And what he's saying is they're moving everything at Google to be deep learning based, including the core of their business.

So the context for Smartvid actually goes all the way back-- and also the context for this project, goes all the way back to when I was first getting into the industry in 2005. So I went to a conference here in the New York City yacht club. Has anybody been there? Anyone been there?

It's amazing. It's just like beautiful. It was-- you know, there's replicas of yachts of the membership all over the walls. It's a really amazing facility. And there was a conference there on AEC technologies.

So I'm pretty new to the industry. It's 2005. I thought, wow, this industry, they must spend a lot of money on technology. Look at this place. It's amazing and they must really focus on. It because I also heard while I was there the CEO of a large EPC contractor, I think it was Fluor talking about how they had just spent $20 million on a cross project risk management system that was integrating in cost and schedule data to try and predict which projects were going to be at risk of having substantial delivery problems.

So this beautiful place, and then the fact that this large company was spending so much money on risk production, made a really big impression on me. I saw that there was an opportunity there. And I saw that there's something to it around risk in our industry.

Then I got a little bit of a cold splash of reality as I went out into the field and started talking with firms and saw that Fluor and some of these other EPC firms were really not indicative of the market as a whole. That as a whole, construction firms don't really spend a lot of money as a percentage of total revenue on technology. And so making a big capital outlay, like $20 million, is really-- it's rare. You won't see that very often.

But the core lesson remained true, is that the traditional drivers of risk, quality, cost, and schedule, which I have here is gears, because you can kind of dial some up and down. You know, it's the classic fast, cheaper, good, pick two kind of dynamic. But then there are some underneath, whether it's safety, and then after working with Suffolk we've also added experience there, those are bedrocks that are unchangeable.

So what we've done at Smartvid is we've focused on safety to start. And we have some work that we're doing, like in the quality arena, which was actually highlighted today on the main stage. So the folks who were on the main stage maybe you remember, the McDonald guys, Corey talked about Smartvid being used to identify quality defects in the tunnel walls automatically using computer vision. So that's how our stuff works.

I've just got a couple more slides of background on the company and what's going on. And then we'll get into the results. What we've basically built is a photo and video management platform that pulls in data from various sources. Think of it as visual content and other content. And then uses an AI engine to analyze it. And then lets you do stuff with the results.

So the lets you do stuff part is where it gets interesting. We have automated reports that can generate overall risk indexes based on what are AI, which we've nicknamed Vinny after one of our first users. So IBM has Watson. We have Vinny. Vinny can see things. He can hear things. And then we aggregate them into data you can use to make better decisions.

As of this point, Vinny has been trained on over three million images. It's actually over 3 and 1/2 million images now. So Vinny is slowly but surely getting smarter. He can see more things. He can make better observations.

So he does this through a capability we call safety monitoring. At the core AI piece is Vinny seeing things and categorizing them. But there's a lot that goes into just having an AI engine versus having a full production grade system that you can plug into BIM 360 Field, or Ox Blue, or Box, whatever your source of data is. And then Vinny is going to start crunching on the data.

And that's what we've been working on with Suffolk over the last couple of years. So getting Vinny up and running for observations. We're using that word a lot. It's almost as much as we're using the word smart. So using the word observations as a way of talking about what Vinny was doing to start.

So this is real data from Suffolk showing the flow of photos from these various systems into Vinny's engine. And then this is a snapshot of some of the data that Vinny has been generating for Suffolk today. So looking at-- these are a number of photos. How many people are seen in them. How much compliance issues.

So this would be things that Vinny can see. Like, here's a worker, and that worker is missing a hard hat, or that worker's missing gloves. That would qualify as a non-compliance issue. And then starting to generate statistics over time.

Now these kinds of reports get at the questions that a lot of people have, which is basically to start, where are the photos coming from? Do we have enough photos? Could we actually see anything in these photos? So through our automatic integration with BIM 360 or your other sources of data, what we try and do is just say, hey, let's not wonder. Let's see. Let's just connect up, and let's run some of these reports and actually see what Vinny can see.

It lets you also do stuff like this, which is compare multiple projects with each other across these categories of compliance. It's been a huge area of emphasis for us over the last six months, is to iterate on these reports themselves. So Vinny is seeing things, person, missing hardhat, other stuff, which I'll show in a second. And then what we're doing is we're building on top of it a reporting infrastructure to take that and make it actionable.

So this is one of the first things we did. This is in collaboration actually with the environmental health and safety team at Suffolk, as well as some of our other customers. We actually start to color code based on how many-- there's actually some sick Sigma concepts in here-- how many standard deviations you are away from the average in terms of-- as defined as what good-- sort of the average of good performers. So that you can flag projects that are falling behind in red and yellow, and highlight the ones that are positive. You can also compare yourself against other projects that are actually in your portfolio.

We've also added trend date. If you're interested in this, by the way, you're welcome to come up afterwards or come to the booth. I'm not going to spend a ton of time on it, because it's more context. This is all the observation type work that we've been doing with Vinny.

So what Vinny can do is see, for example, this is-- Suffolk has a 100% glove compliance policy. Very hard to keep an eye on it. But if you have Vinny, which is an automated eye watching all the time, you're actually better able to get data like this. Like across all jobs here's the percentage compliance by week. And then for each of these jobs you start seeing things like this. Like, the 72% out. That 82% jumps out. Those are places where you may want to pay some attention, because they're trending in a direction where they may need some intervention.

But Vinny is not just looking at PPE. There's actually a whole bunch of, what we would call, leading indicators of risk, right? So Vinny can see a whole bunch of stuff now. Everything from, is the job a mess? So housekeeping issues. A whole category of risks around slip, trip, and fall hazards. So we just rolled out, for example, Vinny's ability to see standing water, ladders, ladders of different types, which was actually another common request from the health and safety teams, because even if you have a ladders last policy, like Suffolk or Skanska. Sometimes it doesn't-- it really does matter rather which kind of ladder you're using. So you want to know what's going on in the field as an indicator-- again, a predictive indicator of risk. Question?

AUDIENCE: What is the source of the images.

JOSH KANNER: Yeah. So the source of the images is a whole bunch of systems that you may have today. So if you're using BIM 360 Field, if you're using some other system that I can't talk about.

AUDIENCE: [INAUDIBLE].

JOSH KANNER: Yeah. So it's basically-- it's your progr--

AUDIENCE: It's a camera system [INAUDIBLE].

JOSH KANNER: Well, you can use-- so one of the things that we do is we let you use any system to gather the photos. So Ox Blue is a camera system. So it takes photos at a certain interval. And those are all fed into the analytics. Yeah. Question.

AUDIENCE: [INAUDIBLE] do you guys also process [INAUDIBLE].

JOSH KANNER: So we support 360 images, too. We actually just announced that this week. So if you're doing spherical data capture, Vinny-- and if you come by our booth I can show you a demo. It's pretty cool. Vinny can spot people and spot other things and hazards in a sphere, and then whoosh, rotate to it and show you where it is. Yes. Question.

AUDIENCE: [INAUDIBLE] GDPR [INAUDIBLE] take photos of people or faces?

JOSH KANNER: Great question. Yeah, so it was a question with GDPR. We actually just did a GDPR audit this summer. And we were found to be very low risk for GDPR issues because of two things. One, in the system itself, we only capture first name, last name, and email of the users. And then the photos are not in any way trying to actually identify who the person is. There's no attempt to identify them. And most of the photos, as you'll see in some of the examples, are kind of from the side or from the back. It's quite hard to tell who the people are.

AUDIENCE: [INAUDIBLE]

JOSH KANNER: Yeah, we haven't we haven't had a real need to blur faces yet. Yeah. It hasn't come up. Any other questions. Yeah?

AUDIENCE: [INAUDIBLE].

JOSH KANNER: Yeah. So will people get fired for this? I hope not. I mean, the whole point of this is actually to move towards driving a positive culture around competing towards compliance, as opposed to trying to nail people for not wearing gloves. You know so that's the way, and this is actually in concert with the health and safety teams that the customers were working with. That's the goal. And that's why we focus on these kind of metrics, as opposed to, hey, here's Bill, and go yell at him. Yeah. Question?

AUDIENCE: So on your list of applications that you're already integrating with, why haven't [INAUDIBLE]. So do you have API or do we have automatically push our photos into your system, and then extract the compliance data out of your system?

JOSH KANNER: Yes.

AUDIENCE: [INAUDIBLE]?

JOSH KANNER: We do. So we-- our whole application is actually API based. So our mobile app and our web app talk to the back end through the same API layer. If you want to use it, we can talk about how you could use it.

AUDIENCE: Are they public, though?

JOSH KANNER: They're not public. No. So if-- that's what I was kind of getting to. If you want to use it, we can talk about how we could set up a relationship for you to do that. And depending on what kind of system it is we might build our own integration adapter so that anyone can use it. If you sign into our system today you can actually-- you can integrate to any one of these systems and more. They also-- we also integrate with Box, with Ignite, with a whole bunch of these systems. You just click on a button, type in your credentials. We don't store them. We just pass a token back, and then the integration starts.

Yeah, sure. I love the questions keeping you guys involved. We'll keep it going, because I don't want to steal too much of Nick's time about the good stuff. But please feel free to-- after I said that, please feel free to ask questions as we go.

So another question-- so the question is always out there, where do the photos come from? Thank you for asking that. Another question that always comes up is, well, what does Vinny see? And what can Vinny do? So here's some additional examples. There's like another 12 or 13 things that Vinny can see, but these are just some examples from our work with Suffolk.

But to the question of do people get fired? There is a photo behind every data point. So what's been interesting is there's actually a really high correlation between Vinny finding stuff that is worth looking at from a PPE standpoint or other risk factor standpoint, and then other stuff that's just going south in that particular photo or that snapshot from that job. So this is an example from Suffolk actually, which it's OK to share. We did it in a webinar together. It was me and the regional head of safety, Marty [INAUDIBLE].

In this image-- it's actually kind of a cool example. So Vinny found this guy, highlighted him because he's missing gloves. So from a machine learning standpoint, this is actually a pretty challenging problem. The hand is really small. As a percentage of the total pixels in the image it's actually very small.

Also, image recognition as a whole, as a discipline, has really been geared around consumer uses. So cats on sofas, or finding photos that have a birthday cake in them. Where typically in consumer photos the object of interest is really pretty close to the center or near the center, construction photos, this is actually zoomed in of a much larger shot where someone was taking a progress photo of what was going on in the deck as they were doing a basically some walk before port.

And they found this guy up on the scaffold on the side. In addition to not wearing gloves, this ladder is improperly placed. IT doesn't have enough actually extension over the top of what should be a platform, but it's not. It's a single plank. And he's not tied off.

So each of these images actually reveals a whole bunch of stuff that's actionable. And then within our system you can take action on it with the project manager, or, and this is another thing that we're announcing, you can actually create an issue in BIM 360 Field. And the imagery and the tags from Smartvid will go automatically into the whole issue workflow in BIM 360 Field. So you can create a safety observation and actually pass it all through into your workflow.

AUDIENCE: Can it automatically create [INAUDIBLE].

JOSH KANNER: We haven't gotten to the automatic issue creation yet, because most of the-- we encourage you to look at the stuff before you do that. Because you may want to add a description. You may want to kind of figure out who you want to do it with.

We have for the reports in that data I showed, we've created an automatic report generation which has been a very common request. So every morning you'll get in your inbox-- or maybe it's every night like at the close of business-- you get in your inbox what happened the day before for the things that Vinnie can look for that you're interested in. So you get like a rolling kind of status update of what Vinnie has seen and what you may want to keep an eye out for without having to go in the system. Yeah, question?

AUDIENCE: For the ladder situation, for example, [INAUDIBLE] is there any thought of doing almost instantaneous UIs?

JOSH KANNER: Yeah.

AUDIENCE: So the actual users then be like, oh, actually I can't go [INAUDIBLE] because it's not [INAUDIBLE].

JOSH KANNER: Yep, so stay tuned. We're actually going to get to that later in the presentation. Everything I've shown you as of right now, this is in the product today. You can do it right now-- the reporting, the automated reporting, all this stuff is there right now. What we're about to move into is the project that we worked on together using data to move-- because as you guys have seen and I can see the interest level, this is actually pretty cool. I mean, it's pretty powerful and it's always working, it's always on, it's looking at all the photos that are coming through those systems. But ultimately, if you take a broader view of what's going on in the industry as a whole, observing through AI is really just the beginning.

NICHOLAS CARBONE: And that's kind of what we challenged Smartvid with when we sat down initially. After seeing the success of this product, I mean, it's really exciting. It's a great idea to use machine learning on images in order to identify safety trends because images don't lie. I mean, they're starting to lie now with new technology, but something like this doesn't lie.

And that's where we sort of sat down together and said, well, can we take-- how do we move from observing to predicting? How do we move from as this worker's missing gloves to this project is going to have an incident? Because as we've discussed a little bit, we can identify the fact that the worker wasn't using gloves, but there are a host of other issues you might find on the project through these images that Vinnie can't identify. So the real thought is that there is a correlation between compliance to something as simple as PPE and actual safety behavior on the project itself.

JOSH KANNER: One way to think of it is if there's a bunch of things that happen on a job that all lead to a potential incident, you can think of those things as being signal. There's noise and signal, and if you can understand what all the-- get a bunch of different signals in and then try and figure out, can you predict that specific outcome from that signal. And everybody knows, actually people sometimes turn up their nose at PPE, but PPE is the number one factor that environmental health and safety professionals look at when asked what is the most predictive indicator of a job site having problems. It's actually data from Dodge smart market report.

NICHOLAS CARBONE: Exactly. So we sat down and we challenged ourselves to that because construction is a great-- because we knew for a fact that a lot of different industries are already in the predictive space. Hospitals are predicting post-surgical infections as well as patient falls. Chevron is actually predicting maintenance before it's needed so that it can do more proactive maintenance on their equipment. UPS is using predictive analysts to more exactly identify when every single one of their packages will arrive. Quad metrics can predict with about 90% accuracy whether a company is going to have a data breach, and internet sales companies all over the world are using predictive analytics daily to manage their user churn as well as to target their marketing dollars.

So the key in each of these different industries is centering the predictive analytics around a single core concept-- patients, equipment, packages, computer systems, and users. So Suffolk, and the construction industry in general, centers itself upon the core concept of a project. But centering on a project sounds very simple and it's certainly not a novel concept, but truly putting it into practice is surprisingly tough. Because as many of you know, every project has a huge number of disparate systems, people, functions, processes, data that all have to be combined and harmonized in some way before you can truly leverage that data. And even then, once you've truly combined it still you need to build a predictive model on top of that in a way that uses it as clean data.

And that's what was really compelling about Smartvid because Smartvid has truly taken the truism that a picture is worth a thousand words and leveraged it. Photos are a very efficient way of capturing a lot of data, and in conjunction with their machine learning algorithms produce clean data. So we took 700,000 plus images from 360 Suffolk projects over the course of 10 years. Pass that through Vinnie with their help.

We then added additional data from within Suffolk-- project information, trade partner information, weather information, historical incident information, et cetera-- and added that as a second layer to a multilayer image learning-- I'm sorry, multilayer machine learning algorithm in order to try to answer the question-- and this was entirely speculative, we didn't know if this would work-- will an incident happen in the next week? And I'm not talking about in the next week at Suffolk, we're talking about on each individual project can we predict whether an incident will occur? We're doing this now. It's certainly not perfect, but it's an incredibly compelling start and an incredibly compelling result that we're now starting to productionize at Suffolk.

JOSH KANNER: Can I jump in for one sec?

NICHOLAS CARBONE: Yeah, go ahead.

JOSH KANNER: So one of the things that's interesting and I just want to highlight a couple of things about what Nick is saying. So first of all, it was a tremendous amount of data. So it's a decade of data, every single photo from every single project. Also there was other information in there that's important to highlight. It wasn't just photo analysis, it was when was the project start and stop time, what was the weather during the phase of that project, what phase was the project in? Some info about the trade partners, although not as much as we can use going forward.

And then all of the incidents. So if you think of this as a time series, you have a project start and stop. In that project you have a whole bunch of photos being taken in a time series, and then you have an incident here, an incident here, an incident there. It was we won't say the exact number, but thousands and thousands of incidents in that 10 year period. Another statistic I like to use as a way of just getting a hold on how much data it was. If you did every project end to end, it's 180 years worth of constant project data went in to train this Vinnie predictive model.

So that kind of gives you a feel. The other thing that's kind of interesting to think about-- and this was something that [INAUDIBLE] said when we were kicking off the project-- is that other data across that 10 year span changed a lot. There were like three different inspection systems, there was a whole bunch of other data that was in different places. The incident data was all standardized, but a lot of these other systems had come and gone.

So having photos from everything from those systems to also the just standard progress photos-- and actually, I should say that there weren't safety photos in here. So the photos were not from safety incidents themselves. It was progress photos, it was-- because you could say, well, if you more pictures because you have more safety incidents, of course that's going to be correlated with risk. But it actually wasn't that. So only recently has Suffolk started taking photos for safety reasons which would then cause that.

NICHOLAS CARBONE: So this is a sort of summary slide on one project at Suffolk. As we said multiple times, this is real data from a real project, and this is the real output from the algorithm. And we sort of summarized it, this is not a direct output from the algorithm. But what we have here on the x-axis is time broken down by weeks throughout the project, y-axis of the algorithmic probability of an incident, i.e. what the algorithm predicted, the probability of incident would be each of those weeks.

Each of these dots is a predicted incident. The orange dots are false positives that were predicted that never actually happened. The blue dots are true positives, i.e. successful predictions of an incident. And as you can see, as the algorithmic probability increases the accuracy actually goes up.

JOSH KANNER: So another way of saying that is if Vinnie is 75% confident then 75% of the dots there should be blue because it means that that's around the probability that you have it. But down here below, Vinnie's probability is much lower so you should expect to see more orange. You can kind of see the up and to the right. Not every project looked this way, too, in terms of the dots. Sometimes the curve goes like this, but in terms of the probabilities over time it's just-- this job looks like it got riskier as it went on.

NICHOLAS CARBONE: Well, you're predicting it was getting riskier. It wasn't actually getting riskier. So what we can actually do is then theoretically set a probability trigger-- let's say this dotted line at 50%-- and begin to control how best to use this algorithm to actually manage our risk.

So let's take a little bit step back from a single projects to the entire set of test projects. That's about 70 different projects across 1,500 project weeks. And just an aside on machine learning, when I say test portfolio, test set, that's because when you're training a machine learning algorithm, you take all your data and you reserve a portion of it-- usually around 20%, depends entirely on who's doing it, and why, and what the algorithm is, but let's say about 20%. You train your algorithms on the rest of that, on that training data, and that's where you tweak the parameters, that's where you tweak the features, that's where you tweak every aspect of that algorithm until you're confident in its results. And only then do you actually test it against that test set. Because that is a set that the algorithm has never seen, and therefore it's the best stand-in we have for real data flowing in in real time.

So taking a look across this, we actually can see the number of alerts that we at Suffolk would get over the course of a year per project at different confidence triggers. So for example, right here if we set that conference trigger at 81% that means that we would be receiving about four alerts per project per year, three of them actual potential incidents. That's three people that we could potentially help get home safely that night.

But safety is every GC's highest priority. And so let's reduce that confidence a little. Let's say we reduce the confidence trigger right now to 66%. We would be alerted 12 times per project per year, eight of them for actual incidents. That's eight people we could help get home that night. Drop it further, 47%. 14 people we could potentially get-- per project per year could potentially get home safe where they wouldn't otherwise.

And if we truly mobilize for these false positives, is that a bad thing? That project is going to get safer even if we mobilize this because we will have the processes in place and the people looking around, making sure that everything is running smoothly. Getting people on a project is not a bad thing, and having a good reason to do it is an even better thing. And we can then manage as a company both at the floor of risk-- or floor trigger-- and then for each project set a separate trigger a level that we can then manage and evolve with time. And that is how you manage risk going forward.

JOSH KANNER: Yeah, when we showed this to Alex Hall who runs health and safety for Suffolk nationally, he had a couple of really interesting responses. The first was 81%, so that's basically how confident you want Vinnie to be before the alert goes off, right? So he said 81 is way too high. We had 66%, which means two out of three, obviously, times Vinnie's going to be right. That seemed much more reasonable.

And the other thing he said is it's kind of interesting from an ongoing data standpoint. He said, our goal is going to be to prove Vinnie wrong every single time. So that data, if we're successful, then the relationship winds up being--

NICHOLAS CARBONE: You can never prove predictions wrong. So that would be my response to that in the sense of Vinnie's going to say it's there. If we actually prevent it, there's no proof that it was actually going to happen.

JOSH KANNER: Fair enough.

NICHOLAS CARBONE: So we'll have to argue with him on that one.

JOSH KANNER: I'm sure he'll enjoy it.

AUDIENCE: There's a Tom Cruise movie about that.

JOSH KANNER: Yes, we'll talk to the precogs.

NICHOLAS CARBONE: Exactly.

JOSH KANNER: Yes, question?

AUDIENCE: So what did he or say [INAUDIBLE] say-- OK, 2/3 of the time we're going to be right, what's your job going to do-- so if you told me as a project manager you have a 66% chance of an incident, you have a range of responses. Shut the job down so we don't hurt somebody, you're not going to do that. Or flood the job with safety people to avoid everybody [INAUDIBLE] an incident, not financially responsible. So what are the results of 66% [INAUDIBLE]?

NICHOLAS CARBONE: So that is a great question, and that is one that we as an organization are still trying to identify and still working out. Alex is working with his safety team to try to identify what he views and what they view as appropriate remediation for these moments, and that's part of productionizing.

JOSH KANNER: There's two things that are recommended for now. The first one is in our industry these alerts are not happening in a vacuum. It's happening at a certain point of time, and that job is in a certain phase of work-- or at least predominantly so. So there are two things that we're talking about. One is having additional tool box talks and raising awareness on the job that, hey, Vinnie has highlighted this. So that's one.

And two-- and there's actually another thing that we're announcing-- is we've published a case study together and when Alex read the case study he actually was pretty-- it was pretty cool. This doesn't happen often, but as an executive he said I want to add a quote to this thing. Because what he sees as the primary value of this is in him allocating some of the floating safety resources he has, because regional safety managers go to those jobs. So not flooding it with safety folks, but maybe Marty who manages 30 jobs should pay a little bit more attention to this job this week because of what's being seen.

AUDIENCE: So related to the question that the 2/3 probability is not what kind of incident, it's just that an incident in general?

JOSH KANNER: Correct.

NICHOLAS CARBONE: Yes.

AUDIENCE: First of all, protective stuff might say what if you aren't wearing gloves, but you're not saying it's going to be--

JOSH KANNER: No. Nope.

AUDIENCE: You're saying it's a general [INAUDIBLE].

NICHOLAS CARBONE: At the same time as the algorithm and the data gets better as we plug in more data feeds, it'll get more accurate and we can start narrowing in on more the region-- maybe not the area, but the specific type of incident that might occur, or the specific trade that might have it. And that's where it starts to become truly transformed because you're right. It's very difficult to just respond right now and say, well, there's going to be incidents, you know, whatever.

It's a billion dollar project and you've got 2,000 people on the site, what do you do? So we can tune our responses to the project itself, but then as the algorithm gets more accurate, as we incorporate more and more data, we can narrow it down closer and closer. And that's everything that they're talking about that Alex is working on with his team.

AUDIENCE: Would you put any information in there relative to whether they're on track with the schedule or financially [INAUDIBLE] performed well or not so well?

JOSH KANNER: That's a great example of the kind of data that in the next version of this thing we're going to include. There's a ton of other factors that you would call really low hanging fruit-- leading factors that you'd want to put in there. Who specifically are the trade partners and what's their history? Is the job on track or not?

NICHOLAS CARBONE: Like we already have a schedule phase in this. Moving on to more detailed schedule aspects is exactly what we're planning to do.

JOSH KANNER: So I should say, by the way, that we're starting an industry group called the predictive analytics council, of which Suffolk is going to be the chair, where we're inviting other companies that want to participate in moving this whole concept forward can contribute data and be a part of developing the next generation of these models.

AUDIENCE: Eventually are you planning on the opposite [INAUDIBLE], like live stream 360 views within speaker systems embedded into it that you can put around your job site that would alert anyone that says-- that maybe would just say, hey, watch the ladder kind of situations?

JOSH KANNER: I don't think so.

NICHOLAS CARBONE: I mean, that's a great vision. I think we're a lot farther away from that than--

AUDIENCE: Yeah, because 360 video imagery involves a huge amount of data that's actually interpreted.

NICHOLAS CARBONE: Yeah, I think that we're very far from that.

JOSH KANNER: I think there's also-- I mean, part of this is going back to the Tom Cruise Minority Report reference. Part of this is exactly the issue of what's the best way to lower incidents, right? And also have a workforce that's engaged and isn't really put out and oppressed by this. And at the same time, you know can you have a system that is-- part of this is the accuracy of it. So you don't want to be barking at someone through a speaker for something that's wrong.

AUDIENCE: But I would address that and say even if you could, I'm not sure I would. Because I would suggest you get a greater impact on dropping incidents if you change safety culture rather than saying, I see you don't have gloves on, go put gloves on. Because imagine the cost it would take to have speakers and gloves everywhere telling the guy without gloves on, let's just make sure that everybody knows you have to have gloves on or we're going to be a project that's on [INAUDIBLE].

NICHOLAS CARBONE: Exactly. And that punitive aspect of it, going back to are you going to fire people, should never be part of this. Because when people feel that 1984-esque monitoring, no one's going to behave in best practices because they're going to be very, very-- no one's going to behave the way they should because no one likes to be monitored that way.

JOSH KANNER: And I would think as a whole Nick and I-- I'll speak for Nick and [INAUDIBLE]-- we're all figuring this out. And there is a lot of, I think, exciting things for us to work on together as an industry about what's the right way to harness these kinds of predictions to ultimately improve safety culture and the other dimensions that will help you reduce incident rates.

NICHOLAS CARBONE: And this is a great way to help people partner with us on the safety culture in the same way that in my view data transparency is almost never a bad thing. Where as long as you present the right data to the right person, no matter what it tells you, whatever the interpretation is, as long as the response is not punitive we can improve it.

You show that a project's doing poorly, let's say, to an executive, if they go back and start yelling at their subordinates nothing's going to get fixed anyway. If they instead can go to people and say, this doesn't look good, what can we do to improve it? The projects are going to go better, and the same way for safety.

If you go and say, well, these are the issues, how can you help us improve it? Everything's going to go much more smoothly, and that creates the true safety culture. So if you had a quick question, I should probably-- there's a few more slides.

AUDIENCE: Does Vinnie do a prediction based on one incident, for example, standing correctly on the scaffold so it'll predict that they fall, for multiple incidents that happen simultaneously in different places, or multiple sequenced actions, for example, [INAUDIBLE]

NICHOLAS CARBONE: So it's a little bit of all that, and I would say Vinnie gets sort of extracts out to some degree the culture of safety much more than specific observation. It says-- and correct me if I'm wrong-- it says more something along the lines of across this project today I saw 50% fewer issues than I did yesterday. How does that affect the safety, the predictive analytics score? It's much more broad than tracking from specific observations because any machine learning algorithm having a single data point, a single photo, a single set of clean data from one moment in time is not going to be very useful to actually predicting what's going to happen at any other point in time.

JOSH KANNER: Yeah, one way I would describe it is-- so you're thinking of it linearly. One way to think of it is it's like a pot of stew really. And you're throwing in all these different things. So Vinnie's observations of safety are one, another thing is the weather, another thing is the project phase. We're not putting in budget and schedule yet, but we'll add that. You know, it's like a pot of stew, and you're stewing it around, and every so often that flag goes off because something in the pot of stew has been seen that across the 180 years that when these certain ingredients come together the risk is really high. Actually, another innovation in the model that we're working on for the next version is to figure out which of the ingredients actually triggered the alert.

NICHOLAS CARBONE: And that gets back to how do you target that remediation?

JOSH KANNER: Because you can actually pull it back. Yeah. So I think behind all of this stuff, though, is-- there's a lot of potential, but there's a lot of challenges. So is this all really worth it at the end of the day? Segue to your next slide.

NICHOLAS CARBONE: Really quickly, on the next-- I'll try to keep this short because there's definitely some questions in the back. So let's imagine what this will actually work-- so we've been doing this on a limited project basis. Now let's think about what the impact would be rolling it out across the company.

So let's pretend for a conservative assumption that we print only a quarter of the incidents that are accurately flagged. And let's also say we are setting our confidence trigger really, really high. Well, across the entirety of our portfolio that would avoid 40 incidents per year. It's 40 people coming home safely. But you know, that's combining a really conservative estimate of remediation with a very, very high confidence trigger. So let's lower the confidence to what, say, Alex Hall said and say 66%. That would prevent 100 incidents per year at Suffolk.

Now as we've hinted at and talked about a bit, as we incorporate more data, more data flows, plug this into more projects, and start incorporating more information, algorithms are only going to get better. So the estimate in that one quarter of the accurately predicted incidents are going to be prevented starts moving from a conservative estimate to something that's ludicrously low. And so that's the truly transformative power. As this continues to evolve with time, the number of incidents fractionally that we can prevent are going to just keep going up. And we're going to do a better and better job, and we can actually drive and manage risks on our projects better.

And this does save money. Yes, it really does. But at the same time, that's not what's important about it. What's important about it is it actually saves families, it saves people.

Imagine 100 people on just Suffolk projects across a year. That's enough for every GC, an entire construction industry. Scale it up to the industry. Improve that so you're preventing half of the incidents that are predicted-- 3/4, who knows? This is the truly transformative power of this, and that's what makes it really, really exciting for us at Suffolk and hopefully for many of you in the audience.

JOSH KANNER: Yeah, and it's just the beginning.

NICHOLAS CARBONE: It's just the beginning.

JOSH KANNER: So one of the things that we're working on with Suffolk is adding to what Vinnie sees with what people see. So we're putting together think of it as a fully integrated predictive risk platform. That's a lot of buzzwords in there, but basically it's not just Vinnie stuff. It's field observations for safety personnel, and then also Alex is really keen on another whole area of risk, which is the pre-task planning and high hazard analysis process.

So the goal here is over the course of the next three, six months, we're actually going to be building out all underneath you can think of it as this pyramid building up to better predictions where we're improving Vinnie through gathering more data through the predictive analytics council, adding in real observations from both field personnel as well as executive job walks, and then do-- safety walks, thank you-- and then doing PTPs and high hazard analysis which are also really powerful indicators of what kind of things, what tasks are going on in the job, and can guide [? tide ?] toward prediction.

NICHOLAS CARBONE: And that incorporates the expert knowledge of all of our professionals into this algorithm in a way that it's not achievable with photos so that we get the both best of both worlds. The really clean data coming out of photos, which as technology involves in construction the photo streams from all the projects just continuously increase, text increases, speech increases, video increases, and tagging all of those sort of more unbiased sources of data into this combined with the human experts putting their information in. It's going to do nothing but improve the algorithm and nothing but help drive safety across Suffolk and across the industry.

So really quickly if you want to learn more, we're just going to stick around a little bit after this and answer more questions. You can also go to the booth, read the predictive analytics case study, and if you want there's some really nice little socks you can pick up. I'm wearing mine, they feel really nice. So Yeah, any additional questions? You had-- I saw you from before.

AUDIENCE: I just wanted to clarify that the predict incidents, I think you had mentioned that there's no categorization to say that 30% are slip trip falls, or 40% [INAUDIBLE] like a hazard, it's all just grouped?

JOSH KANNER: Yeah. Go ahead.

NICHOLAS CARBONE: Right now it is, but we have the information in the algorithm of what all the historical incidents are. And being able to then link it to what's actually going on the site allows us to then start bringing it in to say it's probably going to be-- it might be a slip, trip, or fall, or it might be linked with this specific trade on the site. So as of now, you're right that it's a relatively broad bucket, but version 2, which is going to be coming out pretty soon for us at least, is-- that's exactly what we're working on because that is what's going to really help the remediation side of things.

AUDIENCE: Yeah, and so a related question just on the cost. I think it's actually really important to be able to get by and say at the business level, say actually to be more-- [INAUDIBLE] these risks here or whatever, you can actually see the value that the company saves by doing that. But I'm curious about how the numbers are determined here if you're not able to separate out what types of incidents you're looking at.

JOSH KANNER: Yeah, so I can take that. So what we did, it's really just sort of rough numbers, but it's actual incident cost data summary from the National Institute of Health cost of occupational injuries in construction in the United States. Which that study is from 2007, so I took those numbers, and grossed them up to 2018 dollars, and got an average cost. It's really an average, but the distribution is really wide in incidents. But the average cost of an incident is around $36,000.

And when I kind of sanity checked that with folks in our customer group everyone's like, yeah, that's about right on average, it could be $1,000 it could be $1 million, but that's sort of where it-- did that answer your question, by the way?

NICHOLAS CARBONE: Yeah. So any other questions in the next four minutes? And otherwise we're happy to let you guys go, and we'll stick around if anyone has some additional questions.

JOSH KANNER: Yeah, come by. I'm putting on my shameless selling hat. Come by the booth and if you have been through 60 field projects, you can integrate them in less than 90 seconds and you can start seeing the Vinnie reports that we're generating today automatically in the next day. It just works and you'll get socks. Thank you for your time.

______
icon-svg-close-thick

Cookieの設定

弊社にとって、お客様のプライバシーを守ることと最適な体験を提供することは、どちらも大変重要です。弊社では、お客様に合わせてカスタマイズした情報を提供し、並びにアプリケーションの開発に役立てることを目的に、本サイトのご利用方法についてのデータを収集しております。

そこで、お客様のデータの収集と使用を許可いただけるかどうかをお答えください。

弊社が利用しているサードパーティのサービスについての説明とプライバシー ステートメントも、併せてご確認ください。

サイト動作に必須:オートデスクのサイトが正常に動作し、お客様へサービスを提供するために必要な機能です

Cookie を有効にすることで、お客様の好みやログイン情報が記録され、このデータに基づき操作に対する応答や、ショッピング カートへの商品追加が最適化されます。

使用感が向上:お客様に最適な情報が表示されます

Cookie を有効にすることで、拡張機能が正常に動作し、サイト表示が個々に合わせてカスタマイズされます。お客様に最適な情報をお届けし、使用感を向上させるためのこうした設定は、オードデスクまたはサードパーティのサービス プロバイダーが行います。 Cookie が無効に設定されている場合、一部またはすべてのサービスをご利用いただけない場合があります。

広告表示をカスタマイズ:お客様に関連する広告が表示されます

Cookie を有効にすることで、サイトのご利用内容やご興味に関するデータが収集され、これに基づきお客様に関連する広告が表示されるなど、効率的な動作が可能になります。また、継続的にデータを収集することで、お客様のご興味にさらに関連する広告を配信することが可能になります。Cookie が無効に設定されている場合、お客様に関連しない広告が表示される可能性があります。

icon-svg-close-thick

サードパーティのサービス

それぞれの情報で弊社が利用しているサードパーティのサービスと、オンラインで収集するお客様のデータの使用方法を詳しく説明いたします。

icon-svg-hide-thick

icon-svg-show-thick

サイト動作に必須:オートデスクのサイトが正常に動作し、お客様へサービスを提供するために必要な機能です

Qualtrics
弊社はQualtricsを利用し、アンケート調査やオンライン フォームを通じてお客様が弊社にフィードバックを提供できるようにしています。アンケートの回答は無作為に選んだお客様にお願いしておりますが、お客様から自発的に弊社にフィードバックを提供することも可能です。データを収集する目的は、アンケートの回答前にお客様がとられた行動を、より正しく理解するためです。収集したデータは、発生していた可能性がある問題のトラブルシューティングに役立てさせていただきます。. Qualtrics プライバシー ポリシー
Akamai mPulse
弊社は、弊社サイトでのお客様の行動に関するデータを収集するために、Akamai mPulseを利用しています。収集する情報には、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID、お客様の Autodesk ID が含まれます。このデータを基にサイトのパフォーマンスを測定したり、オンラインでの操作のしやすさを検証して機能強化に役立てています。併せて高度な解析手法を使用し、メールでのお問い合わせやカスタマー サポート、営業へのお問い合わせで、お客様に最適な体験が提供されるようにしています。. Akamai mPulse プライバシー ポリシー
Digital River
弊社は、弊社サイトでのお客様の行動に関するデータを収集するために、Digital Riverを利用しています。収集する情報には、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID、お客様の Autodesk ID が含まれます。このデータを基にサイトのパフォーマンスを測定したり、オンラインでの操作のしやすさを検証して機能強化に役立てています。併せて高度な解析手法を使用し、メールでのお問い合わせやカスタマー サポート、営業へのお問い合わせで、お客様に最適な体験が提供されるようにしています。. Digital River プライバシー ポリシー
Dynatrace
弊社は、弊社サイトでのお客様の行動に関するデータを収集するために、Dynatraceを利用しています。収集する情報には、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID、お客様の Autodesk ID が含まれます。このデータを基にサイトのパフォーマンスを測定したり、オンラインでの操作のしやすさを検証して機能強化に役立てています。併せて高度な解析手法を使用し、メールでのお問い合わせやカスタマー サポート、営業へのお問い合わせで、お客様に最適な体験が提供されるようにしています。. Dynatrace プライバシー ポリシー
Khoros
弊社は、弊社サイトでのお客様の行動に関するデータを収集するために、Khorosを利用しています。収集する情報には、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID、お客様の Autodesk ID が含まれます。このデータを基にサイトのパフォーマンスを測定したり、オンラインでの操作のしやすさを検証して機能強化に役立てています。併せて高度な解析手法を使用し、メールでのお問い合わせやカスタマー サポート、営業へのお問い合わせで、お客様に最適な体験が提供されるようにしています。. Khoros プライバシー ポリシー
Launch Darkly
弊社は、弊社サイトでのお客様の行動に関するデータを収集するために、Launch Darklyを利用しています。収集する情報には、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID、お客様の Autodesk ID が含まれます。このデータを基にサイトのパフォーマンスを測定したり、オンラインでの操作のしやすさを検証して機能強化に役立てています。併せて高度な解析手法を使用し、メールでのお問い合わせやカスタマー サポート、営業へのお問い合わせで、お客様に最適な体験が提供されるようにしています。. Launch Darkly プライバシー ポリシー
New Relic
弊社は、弊社サイトでのお客様の行動に関するデータを収集するために、New Relicを利用しています。収集する情報には、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID、お客様の Autodesk ID が含まれます。このデータを基にサイトのパフォーマンスを測定したり、オンラインでの操作のしやすさを検証して機能強化に役立てています。併せて高度な解析手法を使用し、メールでのお問い合わせやカスタマー サポート、営業へのお問い合わせで、お客様に最適な体験が提供されるようにしています。. New Relic プライバシー ポリシー
Salesforce Live Agent
弊社は、弊社サイトでのお客様の行動に関するデータを収集するために、Salesforce Live Agentを利用しています。収集する情報には、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID、お客様の Autodesk ID が含まれます。このデータを基にサイトのパフォーマンスを測定したり、オンラインでの操作のしやすさを検証して機能強化に役立てています。併せて高度な解析手法を使用し、メールでのお問い合わせやカスタマー サポート、営業へのお問い合わせで、お客様に最適な体験が提供されるようにしています。. Salesforce Live Agent プライバシー ポリシー
Wistia
弊社は、弊社サイトでのお客様の行動に関するデータを収集するために、Wistiaを利用しています。収集する情報には、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID、お客様の Autodesk ID が含まれます。このデータを基にサイトのパフォーマンスを測定したり、オンラインでの操作のしやすさを検証して機能強化に役立てています。併せて高度な解析手法を使用し、メールでのお問い合わせやカスタマー サポート、営業へのお問い合わせで、お客様に最適な体験が提供されるようにしています。. Wistia プライバシー ポリシー
Tealium
弊社は、弊社サイトでのお客様の行動に関するデータを収集するために、Tealiumを利用しています。データには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID が含まれます。このデータを基にサイトのパフォーマンスを測定したり、オンラインでの操作のしやすさを検証して機能強化に役立てています。併せて高度な解析手法を使用し、メールでのお問い合わせやカスタマー サポート、営業へのお問い合わせで、お客様に最適な体験が提供されるようにしています。. Tealium プライバシー ポリシー<>
Typepad Stats
弊社は、弊社サイトでのお客様の行動に関するデータを収集するために、Typepad Statsを利用しています。収集する情報には、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID、お客様の Autodesk ID が含まれます。このデータを基にサイトのパフォーマンスを測定したり、オンラインでの操作のしやすさを検証して機能強化に役立てています。併せて高度な解析手法を使用し、メールでのお問い合わせやカスタマー サポート、営業へのお問い合わせで、お客様に最適な体験が提供されるようにしています。. Typepad Stats プライバシー ポリシー
Geo Targetly
当社では、Geo Targetly を使用して Web サイトの訪問者を最適な Web ページに誘導し、訪問者のいる場所に応じて調整したコンテンツを提供します。Geo Targetly は、Web サイト訪問者の IP アドレスを使用して、訪問者のデバイスのおおよその位置を特定します。このため、訪問者は (ほとんどの場合) 自分のローカル言語でコンテンツを閲覧できます。Geo Targetly プライバシー ポリシー
SpeedCurve
弊社は、SpeedCurve を使用して、Web ページの読み込み時間と画像、スクリプト、テキストなど後続の要素の応答性を計測することにより、お客様の Web サイト エクスペリエンスのパフォーマンスをモニタリングおよび計測します。SpeedCurve プライバシー ポリシー
Qualified
Qualified is the Autodesk Live Chat agent platform. This platform provides services to allow our customers to communicate in real-time with Autodesk support. We may collect unique ID for specific browser sessions during a chat. Qualified Privacy Policy

icon-svg-hide-thick

icon-svg-show-thick

使用感が向上:お客様に最適な情報が表示されます

Google Optimize
弊社はGoogle Optimizeを利用して、弊社サイトの新機能をテストし、お客様に合わせた方法で機能を使えるようにしています。そのため弊社では、弊社サイトにアクセスしているお客様から、行動に関するデータを収集しています。収集するデータには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID、お客様の Autodesk ID などが含まれます。機能のテストの結果によっては、お客様がご利用のサイトのバージョンが変わったり、サイトにアクセスするユーザの属性に応じて、パーソナライズされたコンテンツが表示されるようになる場合があります。. Google Optimize プライバシー ポリシー
ClickTale
弊社は、弊社サイトをご利用になるお客様が、どこで操作につまづいたかを正しく理解できるよう、ClickTaleを利用しています。弊社ではセッションの記録を基に、ページの要素を含めて、お客様がサイトでどのような操作を行っているかを確認しています。お客様の特定につながる個人情報は非表示にし、収集も行いません。. ClickTale プライバシー ポリシー
OneSignal
弊社は、OneSignalがサポートするサイトに広告を配置するために、OneSignalを利用しています。広告には、OneSignalのデータと、弊社サイトにアクセスしているお客様から弊社が収集する行動に関するデータの両方が使われます。弊社が収集するデータには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID が含まれます。この情報に併せて、OneSignalがお客様から収集したデータを使用する場合があります。OneSignalに提供しているデータを弊社が使用するのは、お客様のデジタル広告体験をより適切にカスタマイズし、関連性の高い広告をお客様に配信するためです。. OneSignal プライバシー ポリシー
Optimizely
弊社はOptimizelyを利用して、弊社サイトの新機能をテストし、お客様に合わせた方法で機能を使えるようにしています。そのため弊社では、弊社サイトにアクセスしているお客様から、行動に関するデータを収集しています。収集するデータには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID、お客様の Autodesk ID などが含まれます。機能のテストの結果によっては、お客様がご利用のサイトのバージョンが変わったり、サイトにアクセスするユーザの属性に応じて、パーソナライズされたコンテンツが表示されるようになる場合があります。. Optimizely プライバシー ポリシー
Amplitude
弊社はAmplitudeを利用して、弊社サイトの新機能をテストし、お客様に合わせた方法で機能を使えるようにしています。そのため弊社では、弊社サイトにアクセスしているお客様から、行動に関するデータを収集しています。収集するデータには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID、お客様の Autodesk ID などが含まれます。機能のテストの結果によっては、お客様がご利用のサイトのバージョンが変わったり、サイトにアクセスするユーザの属性に応じて、パーソナライズされたコンテンツが表示されるようになる場合があります。. Amplitude プライバシー ポリシー
Snowplow
弊社は、弊社サイトでのお客様の行動に関するデータを収集するために、Snowplowを利用しています。収集する情報には、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID、お客様の Autodesk ID が含まれます。このデータを基にサイトのパフォーマンスを測定したり、オンラインでの操作のしやすさを検証して機能強化に役立てています。併せて高度な解析手法を使用し、メールでのお問い合わせやカスタマー サポート、営業へのお問い合わせで、お客様に最適な体験が提供されるようにしています。. Snowplow プライバシー ポリシー
UserVoice
弊社は、弊社サイトでのお客様の行動に関するデータを収集するために、UserVoiceを利用しています。収集する情報には、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID、お客様の Autodesk ID が含まれます。このデータを基にサイトのパフォーマンスを測定したり、オンラインでの操作のしやすさを検証して機能強化に役立てています。併せて高度な解析手法を使用し、メールでのお問い合わせやカスタマー サポート、営業へのお問い合わせで、お客様に最適な体験が提供されるようにしています。. UserVoice プライバシー ポリシー
Clearbit
Clearbit を使用すると、リアルタイムのデータ強化により、お客様に合わせてパーソナライズされた適切なエクスペリエンスを提供できます。弊社が収集するデータには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID が含まれます。Clearbit プライバシー ポリシー
YouTube
YouTube はビデオ共有プラットフォームで、埋め込まれたビデオを当社のウェブ サイトで表示および共有することができます。YouTube は、視聴者のビデオのパフォーマンスの測定値を提供しています。 YouTube 社のプライバシー ポリシー

icon-svg-hide-thick

icon-svg-show-thick

広告表示をカスタマイズ:お客様に関連する広告が表示されます

Adobe Analytics
弊社は、弊社サイトでのお客様の行動に関するデータを収集するために、Adobe Analyticsを利用しています。収集する情報には、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID、お客様の Autodesk ID が含まれます。このデータを基にサイトのパフォーマンスを測定したり、オンラインでの操作のしやすさを検証して機能強化に役立てています。併せて高度な解析手法を使用し、メールでのお問い合わせやカスタマー サポート、営業へのお問い合わせで、お客様に最適な体験が提供されるようにしています。. Adobe Analytics プライバシー ポリシー
Google Analytics (Web Analytics)
弊社は、弊社サイトでのお客様の行動に関するデータを収集するために、Google Analytics (Web Analytics)を利用しています。データには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID が含まれます。このデータを基にサイトのパフォーマンスを測定したり、オンラインでの操作のしやすさを検証して機能強化に役立てています。併せて高度な解析手法を使用し、メールでのお問い合わせやカスタマー サポート、営業へのお問い合わせで、お客様に最適な体験が提供されるようにしています。. Google Analytics (Web Analytics) プライバシー ポリシー<>
Marketo
弊社は、お客様に関連性のあるコンテンツを、適切なタイミングにメールで配信できるよう、Marketoを利用しています。そのため、お客様のオンラインでの行動や、弊社からお送りするメールへの反応について、データを収集しています。収集するデータには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID、メールの開封率、クリックしたリンクなどが含まれます。このデータに、他の収集先から集めたデータを組み合わせ、営業やカスタマー サービスへの満足度を向上させるとともに、高度な解析処理によって、より関連性の高いコンテンツを提供するようにしています。. Marketo プライバシー ポリシー
Doubleclick
弊社は、Doubleclickがサポートするサイトに広告を配置するために、Doubleclickを利用しています。広告には、Doubleclickのデータと、弊社サイトにアクセスしているお客様から弊社が収集する行動に関するデータの両方が使われます。弊社が収集するデータには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID が含まれます。この情報に併せて、Doubleclickがお客様から収集したデータを使用する場合があります。Doubleclickに提供しているデータを弊社が使用するのは、お客様のデジタル広告体験をより適切にカスタマイズし、関連性の高い広告をお客様に配信するためです。. Doubleclick プライバシー ポリシー
HubSpot
弊社は、お客様に関連性のあるコンテンツを、適切なタイミングにメールで配信できるよう、HubSpotを利用しています。そのため、お客様のオンラインでの行動や、弊社からお送りするメールへの反応について、データを収集しています。収集するデータには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID、メールの開封率、クリックしたリンクなどが含まれます。. HubSpot プライバシー ポリシー
Twitter
弊社は、Twitterがサポートするサイトに広告を配置するために、Twitterを利用しています。広告には、Twitterのデータと、弊社サイトにアクセスしているお客様から弊社が収集する行動に関するデータの両方が使われます。弊社が収集するデータには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID が含まれます。この情報に併せて、Twitterがお客様から収集したデータを使用する場合があります。Twitterに提供しているデータを弊社が使用するのは、お客様のデジタル広告体験をより適切にカスタマイズし、関連性の高い広告をお客様に配信するためです。. Twitter プライバシー ポリシー
Facebook
弊社は、Facebookがサポートするサイトに広告を配置するために、Facebookを利用しています。広告には、Facebookのデータと、弊社サイトにアクセスしているお客様から弊社が収集する行動に関するデータの両方が使われます。弊社が収集するデータには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID が含まれます。この情報に併せて、Facebookがお客様から収集したデータを使用する場合があります。Facebookに提供しているデータを弊社が使用するのは、お客様のデジタル広告体験をより適切にカスタマイズし、関連性の高い広告をお客様に配信するためです。. Facebook プライバシー ポリシー
LinkedIn
弊社は、LinkedInがサポートするサイトに広告を配置するために、LinkedInを利用しています。広告には、LinkedInのデータと、弊社サイトにアクセスしているお客様から弊社が収集する行動に関するデータの両方が使われます。弊社が収集するデータには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID が含まれます。この情報に併せて、LinkedInがお客様から収集したデータを使用する場合があります。LinkedInに提供しているデータを弊社が使用するのは、お客様のデジタル広告体験をより適切にカスタマイズし、関連性の高い広告をお客様に配信するためです。. LinkedIn プライバシー ポリシー
Yahoo! Japan
弊社は、Yahoo! Japanがサポートするサイトに広告を配置するために、Yahoo! Japanを利用しています。広告には、Yahoo! Japanのデータと、弊社サイトにアクセスしているお客様から弊社が収集する行動に関するデータの両方が使われます。弊社が収集するデータには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID が含まれます。この情報に併せて、Yahoo! Japanがお客様から収集したデータを使用する場合があります。Yahoo! Japanに提供しているデータを弊社が使用するのは、お客様のデジタル広告体験をより適切にカスタマイズし、関連性の高い広告をお客様に配信するためです。. Yahoo! Japan プライバシー ポリシー
Naver
弊社は、Naverがサポートするサイトに広告を配置するために、Naverを利用しています。広告には、Naverのデータと、弊社サイトにアクセスしているお客様から弊社が収集する行動に関するデータの両方が使われます。弊社が収集するデータには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID が含まれます。この情報に併せて、Naverがお客様から収集したデータを使用する場合があります。Naverに提供しているデータを弊社が使用するのは、お客様のデジタル広告体験をより適切にカスタマイズし、関連性の高い広告をお客様に配信するためです。. Naver プライバシー ポリシー
Quantcast
弊社は、Quantcastがサポートするサイトに広告を配置するために、Quantcastを利用しています。広告には、Quantcastのデータと、弊社サイトにアクセスしているお客様から弊社が収集する行動に関するデータの両方が使われます。弊社が収集するデータには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID が含まれます。この情報に併せて、Quantcastがお客様から収集したデータを使用する場合があります。Quantcastに提供しているデータを弊社が使用するのは、お客様のデジタル広告体験をより適切にカスタマイズし、関連性の高い広告をお客様に配信するためです。. Quantcast プライバシー ポリシー
Call Tracking
弊社は、キャンペーン用にカスタマイズした電話番号を提供するために、Call Trackingを利用しています。カスタマイズした電話番号を使用することで、お客様は弊社の担当者にすぐ連絡できるようになり、弊社はサービスのパフォーマンスをより正確に評価できるようになります。弊社では、提供した電話番号を基に、サイトでのお客様の行動に関するデータを収集する場合があります。. Call Tracking プライバシー ポリシー
Wunderkind
弊社は、Wunderkindがサポートするサイトに広告を配置するために、Wunderkindを利用しています。広告には、Wunderkindのデータと、弊社サイトにアクセスしているお客様から弊社が収集する行動に関するデータの両方が使われます。弊社が収集するデータには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID が含まれます。この情報に併せて、Wunderkindがお客様から収集したデータを使用する場合があります。Wunderkindに提供しているデータを弊社が使用するのは、お客様のデジタル広告体験をより適切にカスタマイズし、関連性の高い広告をお客様に配信するためです。. Wunderkind プライバシー ポリシー
ADC Media
弊社は、ADC Mediaがサポートするサイトに広告を配置するために、ADC Mediaを利用しています。広告には、ADC Mediaのデータと、弊社サイトにアクセスしているお客様から弊社が収集する行動に関するデータの両方が使われます。弊社が収集するデータには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID が含まれます。この情報に併せて、ADC Mediaがお客様から収集したデータを使用する場合があります。ADC Mediaに提供しているデータを弊社が使用するのは、お客様のデジタル広告体験をより適切にカスタマイズし、関連性の高い広告をお客様に配信するためです。. ADC Media プライバシー ポリシー
AgrantSEM
弊社は、AgrantSEMがサポートするサイトに広告を配置するために、AgrantSEMを利用しています。広告には、AgrantSEMのデータと、弊社サイトにアクセスしているお客様から弊社が収集する行動に関するデータの両方が使われます。弊社が収集するデータには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID が含まれます。この情報に併せて、AgrantSEMがお客様から収集したデータを使用する場合があります。AgrantSEMに提供しているデータを弊社が使用するのは、お客様のデジタル広告体験をより適切にカスタマイズし、関連性の高い広告をお客様に配信するためです。. AgrantSEM プライバシー ポリシー
Bidtellect
弊社は、Bidtellectがサポートするサイトに広告を配置するために、Bidtellectを利用しています。広告には、Bidtellectのデータと、弊社サイトにアクセスしているお客様から弊社が収集する行動に関するデータの両方が使われます。弊社が収集するデータには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID が含まれます。この情報に併せて、Bidtellectがお客様から収集したデータを使用する場合があります。Bidtellectに提供しているデータを弊社が使用するのは、お客様のデジタル広告体験をより適切にカスタマイズし、関連性の高い広告をお客様に配信するためです。. Bidtellect プライバシー ポリシー
Bing
弊社は、Bingがサポートするサイトに広告を配置するために、Bingを利用しています。広告には、Bingのデータと、弊社サイトにアクセスしているお客様から弊社が収集する行動に関するデータの両方が使われます。弊社が収集するデータには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID が含まれます。この情報に併せて、Bingがお客様から収集したデータを使用する場合があります。Bingに提供しているデータを弊社が使用するのは、お客様のデジタル広告体験をより適切にカスタマイズし、関連性の高い広告をお客様に配信するためです。. Bing プライバシー ポリシー
G2Crowd
弊社は、G2Crowdがサポートするサイトに広告を配置するために、G2Crowdを利用しています。広告には、G2Crowdのデータと、弊社サイトにアクセスしているお客様から弊社が収集する行動に関するデータの両方が使われます。弊社が収集するデータには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID が含まれます。この情報に併せて、G2Crowdがお客様から収集したデータを使用する場合があります。G2Crowdに提供しているデータを弊社が使用するのは、お客様のデジタル広告体験をより適切にカスタマイズし、関連性の高い広告をお客様に配信するためです。. G2Crowd プライバシー ポリシー
NMPI Display
弊社は、NMPI Displayがサポートするサイトに広告を配置するために、NMPI Displayを利用しています。広告には、NMPI Displayのデータと、弊社サイトにアクセスしているお客様から弊社が収集する行動に関するデータの両方が使われます。弊社が収集するデータには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID が含まれます。この情報に併せて、NMPI Displayがお客様から収集したデータを使用する場合があります。NMPI Displayに提供しているデータを弊社が使用するのは、お客様のデジタル広告体験をより適切にカスタマイズし、関連性の高い広告をお客様に配信するためです。. NMPI Display プライバシー ポリシー
VK
弊社は、VKがサポートするサイトに広告を配置するために、VKを利用しています。広告には、VKのデータと、弊社サイトにアクセスしているお客様から弊社が収集する行動に関するデータの両方が使われます。弊社が収集するデータには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID が含まれます。この情報に併せて、VKがお客様から収集したデータを使用する場合があります。VKに提供しているデータを弊社が使用するのは、お客様のデジタル広告体験をより適切にカスタマイズし、関連性の高い広告をお客様に配信するためです。. VK プライバシー ポリシー
Adobe Target
弊社はAdobe Targetを利用して、弊社サイトの新機能をテストし、お客様に合わせた方法で機能を使えるようにしています。そのため弊社では、弊社サイトにアクセスしているお客様から、行動に関するデータを収集しています。収集するデータには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID、お客様の Autodesk ID などが含まれます。機能のテストの結果によっては、お客様がご利用のサイトのバージョンが変わったり、サイトにアクセスするユーザの属性に応じて、パーソナライズされたコンテンツが表示されるようになる場合があります。. Adobe Target プライバシー ポリシー
Google Analytics (Advertising)
弊社は、Google Analytics (Advertising)がサポートするサイトに広告を配置するために、Google Analytics (Advertising)を利用しています。広告には、Google Analytics (Advertising)のデータと、弊社サイトにアクセスしているお客様から弊社が収集する行動に関するデータの両方が使われます。弊社が収集するデータには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID が含まれます。この情報に併せて、Google Analytics (Advertising)がお客様から収集したデータを使用する場合があります。Google Analytics (Advertising)に提供しているデータを弊社が使用するのは、お客様のデジタル広告体験をより適切にカスタマイズし、関連性の高い広告をお客様に配信するためです。. Google Analytics (Advertising) プライバシー ポリシー
Trendkite
弊社は、Trendkiteがサポートするサイトに広告を配置するために、Trendkiteを利用しています。広告には、Trendkiteのデータと、弊社サイトにアクセスしているお客様から弊社が収集する行動に関するデータの両方が使われます。弊社が収集するデータには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID が含まれます。この情報に併せて、Trendkiteがお客様から収集したデータを使用する場合があります。Trendkiteに提供しているデータを弊社が使用するのは、お客様のデジタル広告体験をより適切にカスタマイズし、関連性の高い広告をお客様に配信するためです。. Trendkite プライバシー ポリシー
Hotjar
弊社は、Hotjarがサポートするサイトに広告を配置するために、Hotjarを利用しています。広告には、Hotjarのデータと、弊社サイトにアクセスしているお客様から弊社が収集する行動に関するデータの両方が使われます。弊社が収集するデータには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID が含まれます。この情報に併せて、Hotjarがお客様から収集したデータを使用する場合があります。Hotjarに提供しているデータを弊社が使用するのは、お客様のデジタル広告体験をより適切にカスタマイズし、関連性の高い広告をお客様に配信するためです。. Hotjar プライバシー ポリシー
6 Sense
弊社は、6 Senseがサポートするサイトに広告を配置するために、6 Senseを利用しています。広告には、6 Senseのデータと、弊社サイトにアクセスしているお客様から弊社が収集する行動に関するデータの両方が使われます。弊社が収集するデータには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID が含まれます。この情報に併せて、6 Senseがお客様から収集したデータを使用する場合があります。6 Senseに提供しているデータを弊社が使用するのは、お客様のデジタル広告体験をより適切にカスタマイズし、関連性の高い広告をお客様に配信するためです。. 6 Sense プライバシー ポリシー
Terminus
弊社は、Terminusがサポートするサイトに広告を配置するために、Terminusを利用しています。広告には、Terminusのデータと、弊社サイトにアクセスしているお客様から弊社が収集する行動に関するデータの両方が使われます。弊社が収集するデータには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID が含まれます。この情報に併せて、Terminusがお客様から収集したデータを使用する場合があります。Terminusに提供しているデータを弊社が使用するのは、お客様のデジタル広告体験をより適切にカスタマイズし、関連性の高い広告をお客様に配信するためです。. Terminus プライバシー ポリシー
StackAdapt
弊社は、StackAdaptがサポートするサイトに広告を配置するために、StackAdaptを利用しています。広告には、StackAdaptのデータと、弊社サイトにアクセスしているお客様から弊社が収集する行動に関するデータの両方が使われます。弊社が収集するデータには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID が含まれます。この情報に併せて、StackAdaptがお客様から収集したデータを使用する場合があります。StackAdaptに提供しているデータを弊社が使用するのは、お客様のデジタル広告体験をより適切にカスタマイズし、関連性の高い広告をお客様に配信するためです。. StackAdapt プライバシー ポリシー
The Trade Desk
弊社は、The Trade Deskがサポートするサイトに広告を配置するために、The Trade Deskを利用しています。広告には、The Trade Deskのデータと、弊社サイトにアクセスしているお客様から弊社が収集する行動に関するデータの両方が使われます。弊社が収集するデータには、お客様がアクセスしたページ、ご利用中の体験版、再生したビデオ、購入した製品やサービス、お客様の IP アドレスまたはデバイスの ID が含まれます。この情報に併せて、The Trade Deskがお客様から収集したデータを使用する場合があります。The Trade Deskに提供しているデータを弊社が使用するのは、お客様のデジタル広告体験をより適切にカスタマイズし、関連性の高い広告をお客様に配信するためです。. The Trade Desk プライバシー ポリシー
RollWorks
We use RollWorks to deploy digital advertising on sites supported by RollWorks. Ads are based on both RollWorks data and behavioral data that we collect while you’re on our sites. The data we collect may include pages you’ve visited, trials you’ve initiated, videos you’ve played, purchases you’ve made, and your IP address or device ID. This information may be combined with data that RollWorks has collected from you. We use the data that we provide to RollWorks to better customize your digital advertising experience and present you with more relevant ads. RollWorks Privacy Policy

オンライン体験の品質向上にぜひご協力ください

オートデスクは、弊社の製品やサービスをご利用いただくお客様に、優れた体験を提供することを目指しています。これまでの画面の各項目で[はい]を選択したお客様については、弊社でデータを収集し、カスタマイズされた体験の提供とアプリケーションの品質向上に役立てさせていただきます。この設定は、プライバシー ステートメントにアクセスすると、いつでも変更できます。

お客様の顧客体験は、お客様が自由に決められます。

オートデスクはお客様のプライバシーを尊重します。オートデスクでは収集したデータを基に、お客様が弊社製品をどのように利用されているのか、お客様が関心を示しそうな情報は何か、オートデスクとの関係をより価値あるものにするには、どのような改善が可能かを理解するよう務めています。

そこで、お客様一人ひとりに合わせた体験を提供するために、お客様のデータを収集し、使用することを許可いただけるかどうかお答えください。

体験をカスタマイズすることのメリットにつきましては、本サイトのプライバシー設定の管理でご確認いただけます。弊社のプライバシー ステートメントでも、選択肢について詳しく説明しております。