How Analytics Is Bringing Insights to Toyota's Factory Projects

TOMOTOSHI JO: Hello, everyone. Our topic is "How Analytics is Bringing Insights to Toyota's Factory Projects." So this is the safe harbor statement. And my name is Tomotoshi Jo. I have a background in information system, BIM, and project management. I support Toyota Production Engineering team members to use Autodesk products such as Autodesk Construction Cloud, AutoCAD, Inventor, Navisworks, Vault, and many more.

I also help different PE departments build integrated factory models for their new equipment installation and plant construction project. About one year ago, I made a promise to my colleagues to attend AU as a speaker. Here I am. It's my first time to attend AU and be a speaker. I would like to say thank you to all people who helped me so far. Next, I would like to hand over to my co-speaker Liang and let him introduce himself.

LIANG GONG: Hello, everyone. This is Leon Gong from Autodesk Consulting. I am a consultant specialized in analytics and automations. Meanwhile, I'm currently an MS candidate at University of Chicago specializing in AI and machine learning. Thanks. Next slide, please?

TOMOTOSHI JO: So in this class, I will talk about the Toyota Way, which drives me to partner with Autodesk consulting team to bring analytics to Toyota Production Engineering, explain the challenge that Toyota Production Engineering is facing, and lastly, discuss how analytics helps Toyota team understand the current situation and make decisions on the next step.

So here are the four key learning objects of the session. I will mainly speak from the business perspective, while Liang will cover the technical side. So I believe many people know about Toyota company, but you may not know or even heard about Toyota Way. It's the core values of Toyota. To me, the number one core value is to drive curiosity. At Toyota, we ask to discover the mechanic behind phenomena. This mindset generates new ideas.

A second, the number 2 core value is to continue the quest for improvement. In Japanese, it is also called as Kaizen. At Toyota, we believe in the nature ability of people to change things for the better. Even improvement regardless of size is valuable. Encouraging both incremental and breakthrough innovative thinking, we seek to evolve with Kaizen, never accepting the status quo.

And the number 3 value is to create room to grow. At Toyota, focusing on what is essential, we eliminate waste and manage our resources carefully to create room to grow. This is the foundation for agility and cultivation of the new ideas for the future. These three core values are the backbone of this case study.

So every organization has its own challenge. Toyota is not an exception. Since 2020, Toyota has had an EBA contract with Autodesk. As you may know, the EBA is a partnership with Autodesk that can include more products and services than the traditional contracts. But there are three major benefits of EBA. First, token flags, second, the enterprise priority support, and third, advisory and implementation service.

With the EBA, it opens the door for all the members at Toyota Motor North America to use any Autodesk product. If Toyota members have ideas to improve the current process or build new workflow, they can request implementation service from Autodesk. And a dedicated consulting team will help to map the process and deploy new product or features.

While everything has two sides, challenge also comes with EBA. Toyota Production Engineering is joint organization with 10 plants across the nation and over 2,000 engineers working inside plants. From the high level, we always need to consider how to maximize the benefits of EBA, how to capture the value from EBA, and how to adjust by the return of investment of EBA.

So at the working level, we also face the challenges to transition to Autodesk Construction Cloud Build. PlanGrid was used to share information, manage the sheets, joins in the Toyota's factory projects. As you know, PlanGrid was acquired by Autodesk in 2018. And later, Autodesk launched ACC Build, which is the next generation solution for field and project management. Because ACC Build is covered under Toyota's EBA, the management team made a decision to generate a move to ACC Build.

However, switching to ACC Build is not easy. First of all, we have to train not only the users inside Toyota, but also the external users, such as general contracts and subcontracts. Different users have different purpose to use ACC Build. For example, project lead needs to set up tools for the project members in ACC Build, while our project members may just need to upload or download files, create issues, RFIs, and submittals.

Second, we have to update the existing Shikumi. Shikumi is a Japanese word which means "operation procedure." PlanGrid is embedded into many existing Shikumis to move to ACC Build. We must fully understand the whole operation procedure and the function difference between ACC Build and PlanGrid so that we can revise the process to match the purpose of each Shikumi and make it useful to the end user.

And lastly, we have to upgrade the existing dashboard. The data sets of existing Power BI dashboard are imported from Excel files. Those Excel files, sheets contains many formulas. And to switch to ACC Build, we must study the current data set and develop a new one leveraging the data set imported directly from ACC Build.

So here is a glance at current ACC Build usage. We started to use ACC Build in 2021. Currently, there are over 500 active users, 150 projects managed inside ACC. And some are small projects like installing the charging station. Some are large projects like the new Battery Plant Project. There are over 1,000 issues, 3,000 RFIs, and 4,000 submittals created and managed in ACC Build.

So another challenge is to optimize token usage at Toyota Production Engineering. Token flag is a type of licensing model provided by Autodesk and the EBA. It lets Toyota pre-purchase tokens to access any product via daily rate. For example, a user will be charged a fixed rate even just using AutoCAD for 10 minutes in a day. But Toyota has a limited token to use for the whole organization.

Every token comes with a price. We have to watch the token usage carefully, understand how our users use the Autodesk products, and guide them to effectively use those tools. Maybe even suggest them to use some free Autodesk tools so that we could save some tokens and create room for the future growing usage. So under Toyota's EBA, the token usage is categorized into three types-- desktop product, cloud system, and adjustments.

Here, I just want to explain a little bit about adjustments. Adjustments refers to the Autodesk products that are not valuable as token flags. Basically, they charge monthly at a fixed rate. And since May this year, the Autodesk desktop product consumed over 300,000 tokens. The most used products are AutoCAD Inventor and VRED. And the cloud system consumed over 100,000 tokens.

The most used products are ACC Build, being Collaboration and Docs. And for the adjustment, it consumed over 70,000 tokens. And the most used products at Toyota are ACC Connect, ProEst and Pype. To obtain insights of those construction projects and token usage managed in ACC, the DB analyst is required. And I would like to let Liang talk about his strategy and the way to map the analyst's process. Thank you. Liang?

LIANG GONG: Thanks, Tommy. Before we go into the deep sea of the technologies associated with the methodology that Tommy was talking about, I'd like to give the overview picture of where we are, where here the zone is at in perspective, the data strategy approach. So this is a normal evolution of the data strategies for the AEC industry. You could see we start with descriptive analytics, and it goes into diagnostic, which means here, if you're trying to benchmark or scoring the different entities like the projects under your BIM 360 or ACC hub, that's a diagnostic analytics example.

And then we evolve into predictive and prescriptive. And the later two phases, they are more like associated with machine learning and AI, which are hot topics these days. But in order to evolve to the latter two phases, it's always better to build the foundation of the "how solid," which is the descriptive and diagnostic, which is also associated with the database foundation, which I'm going to talk about in the next page. Next page, please?

As you can see here, when we are mentioning the CDE, the connected data environment or common data environments, what exactly they are. In the previous slides, he was talking about the different kinds of analytics during the evolution process. And the foundation of that is really the database. As you can see on this slide on the left side and the right side-- so on the left side, it's more like mimicking the database for all the normalized tables.

On the very left side, the very left column like cost, operations, sketches, design, all those data are siloed data which contain a lot of the normalized table. And on the right side, very right, like the prediction, forecasting, correlations, training, diagnosing, these are the visualizations that are ready to be consumed by the end users. For the end users, they do not need to understand the back end, which on the left side, how the data engineering process is looking like.

They just need to get ready to consume the data, interpret the data for their business purpose. So that's why this slide is divided into two big parts. The first part is data storage environment, which contains all the raw data in silos. And on the right side, the data analytics environment, which is more on the front end like Power BI or Tableau, which are ready to consume the data for the end users.

As you can see here, the biggest problem is that in order to build this data pipeline, we need to perform a lot of the data engineering work because inevitably, there are a lot of silos in our business today. And we need to do a lot of data engineering work to consolidate data, to consume the data in order for them ready to be used for the end users. So basically, using an analogy, you want to build the foundation of the house very solid before you're building the upper structures and the facade of the house. So that's the analogy here.

And next page, I'm going to talk about the data connector. So remember, the left side of this page, consolidation process, is more where the data storage environment is. And in our example-- next page, please. In the ACC in the Autodesk example, the data storage environment is the data connector, which the full name is ACC Autodesk Construction Cloud data connector. If you use network product, if you go to the insight module of ACC or BIM 360, there is a sector called Data Connector.

I put the link here for your reference if you're interested in knowing more about it, those two links, and read the articles. So what it basically does is that all the data you put onto our user interface, UI, onto BIM 360 or ACC, let's say you put a lot of data on our issues module, all those data are going to be organized and put under this data connector ready for you to download and ready for you to consume.

And on the right side, this is how the data connector looks like. It consumes all the normalized table. By normalized table, this is what I mean-- all the different CSVs for the siloed modules on the right side. So this is our use case for the data storage environment under the Autodesk ACC's perspective under this structure.

Next page, I will have Tommy talking about the data analytics environment because I just talked about the data storage environment, which is kind of the backend of this design workflow. And next, Tommy is going to talk about the data analytics environment, which is going to show you the videos that are ready to be consumed by the business partners. Tommy, please take it over for the data analytics environment. Thanks.

TOMOTOSHI JO: So we use the data connector to import ACC Build project data into Power BI dashboard. And here, I just want to discuss about what dashboard we develop and how those dashboard, our analysts benefit toward the production engineering. So first, executive overview. This dashboard shows a summary of the audit projects in the Toyota ACC hub.

The management team could easily view the project start date, project location, number of companies, and members for each project and understand what's happening and going on with inside organization, Toyota PE organization. Second, the Issue Analysis dashboard. So the Issue Analysis dashboard contains the performance metrics such as the average days to close, number of open issues, status of issues in each project. It could quickly help senior managers identify the road block for each project and common issues among all the projects.

So the third is the Forms Analysis dashboard. So at Toyota, compound is used to confirm the quality of all aspects of construction projects. So compound literally means signal. In ACC Build, we use forms to implement this concept. The form dashboard helps the management know the progress and the lead time to complete quality confirmation for all projects. That's really helpful, especially safety is the biggest concern for the manufacturing plant.

Next, the fourth one is the Assets Analysis dashboard. The assets dashboard summarizes all the assets for each project. It helps the operation team understand what equipment will be handed over to them after the project ends and what the status of each equipment is so that they can plan the maintenance in the future. So that's all the dashboard analysis we develop with the data connector.

Next, so here are-- previously, I mentioned the challenge to optimize the token usage at Toyota. Next, I will let Liang talk about ACC Connect and how he developed the dashboard to analyze token usage for us.

LIANG GONG: Thanks, Tommy. So in order to analyze the tokens usage, the first step is to really get the data, the tokens consumption data first before we analyze them, before we visualize them, right? So the first step is how we actually get the data. That really relies on our APS APIs. So two parts-- what is APS? Autodesk Platform Services. It's a cloud service which contains a lot of APIs.

And then what is the API? API is an application programming interface. It is a way for more computer programs to communicate with each other. It is a type of software interface offering a service to other pieces of software. And what is API documentation? It is a document or standard that describes how to build or use such a connection or interface. That is what? An API specification.

So together, this is APS API Autodesk Platform Services that provides an application programming interface for different softwares to talk with each other, including for the software that Autodesk provides to talk to third-party software, external software like shown on the screen like SharePoint, Google Sheets, or DocuSign. On the left side, these are on the slide, which Tom is presenting. On the left side, these are the products that Autodesk Construction Cloud provide, like the Autodesk Build, Autodesk Takeoff, Autodesk Docs.

If I wanted this software, this platform to talk with external software like SharePoint, if you want to have any interactions or automation workflows set up, you need to rely on our APIs, which is the bottom right side pop-up, Autodesk Platform Services, which includes the APIs for the different modules, like for issues, RFIs, et cetera.

And to give a little bit more about what is ACC Connect, ACC Connect is kind of similar to Power Automate, but it's different because it is designed specifically for the Autodesk ecosystem. If you go to next slide, please? So how do we use those APS APIs? We leverage ACC Connect to write those APIs. And regarding ACC Connect, previously-- its parent company is called Workato, but Autodesk rebranded and add our own customized connections and give it a new name, ACC Connect.

What are the use cases for ACC Connect? We see a lot of this usage areas like document management, between DocuSign, between Box, between SharePoint. We're also seeing a growing area for project management systems like Excel, Smartsheet, Google Sheets, how you analyze this. These are more associated with analytics. And another biggest area we see lies under accounting. If you want, for example, want your ACC cost module to talk with the external accounting system, like QuickBooks, how we can automate that workflow, it's going to utilize the APS APIs and ACC Connect.

And in our case here at Toyota, Tommy wants to analyze the token's consumption, we use ACC Connect and APS APIs to extract the data to set up a data pipeline. Next slide, please? So I'd like to give you another example of the application of ACC Connect because it's not only restricted to extracting the tokens's data. Here's another example I like to illustrate.

Everyone's like a lot of us, who work on the construction side here, and we have a lot of QA/QC work to do. In this real example that Toyota wants to create QC and commissioning their equipment on the side in the factory, in order to do that, they wanted to use the ACC Build app. But how do you scan each equipment? So here brings up the concept of barcode. So ideally, we want to put a barcode on each equipment and scan the barcode with the app to bring up all the associated asset and associated Kanban forms with that specific asset.

If we print out the barcode for each asset equipment, it's going to be very time consuming because there could be more than 500, more than 1,000 assets. So in order to automate this workflow in order to save time, we use ACC Connect to automatically generate a barcode column for each asset. And the barcodes are all unique. Meanwhile, we'll print out a PNG file as you can see on the lower right side of the slide and put it under the DOCX file.

So in this way, after the automating process, we could print out a barcode for each asset and stick it to each equipment in the factory. So if you're a QC commissioner, you could just open up your ACC Build app and scan the barcode. The corresponding asset is going to pop up. You can see its associated its own asset information, associated Kanban forms information, associated issues. It's all digitalized. So that's the benefit of automating this workflow with ACC Connect.

Coming back to the topic of token assumptions after we're automatically extracting the tokens data. And now Tom is ready to consume them to visualize these tokens consumption data for his business case. So next page, Tommy is going to talking about the visualization and data analytics for the tokens consumption data.

TOMOTOSHI JO: OK, thank you, Liang. So here, I want to show two examples of dashboard we developed for token usage. The first one is the User Token Analysis dashboard. So this dashboard helps management know the trade of a user account. And based on the user's ID, we could understand which department they belong to and which area they may focus on. For example, the plant they are designed, or the tooling design, or the simulation side.

Next is just another-- the other dashboard is called Product Token Analysis. So this dashboard helps management understand the token consumed and hour used per product. For example, on the right side, you can see the most used product at Toyota is AutoCAD. And since May this year, Toyota engineers already spent over 90,000 hours on the Autodesk CAD product only. So it may help us to consider why and how we could help our users maybe use AutoCAD.

So that's the two examples of token usage analyst dashboard. And in the next here in the end, we would like to conclude this case study and give the recommendation. First for the conclusion-- so by analyzing the data obtained while the data connector and ACC Connect, we can learn project overall performance, individual project members workload, and token usage at Toyota. So it leads us to make the improvement to project planning and management of balanced workload for each project members, and target the people to provide them the right tools and training. So next, Liang will give his recommendation.

LIANG GONG: Thanks, Tommy. So a very important thing, as you probably have already noticed earlier in this presentation that we want to build a solid foundation for the data pipeline for the different kinds of analytics. So right now, as you could see in the chart, we bring the data directly from ACC Construction Cloud into Power BI directly for analytics.

However, when the data are growing more and more, Power BI is going to lose its efficiency because Power BI is not really a data storage tool. That's why we're adding a semantic layer, which is the data house or data warehouse or data storage like a SQL database or Snowflake between Power BI and our ACC Construction Cloud.

The benefit of doing that is listed below for the seven points. I'm not going to read one by one. But overall, you could perform the data engineering, writing the queries instead of the data warehouse before you bring the consolidated, ready-to-consume data sets into Power BI for visualization purpose only. So this going to save tons of time to write the queries in Power BI because that's only going to slow down the performance of Power BI when we're having more and more projects data. So adding the semantic layer of data warehouse here is very important. That's a recommendation to all the audience here.

And the next slide is for predictive analytics. For the first two parts, we're talking about the descriptive and diagnostic analytics. And now I like to talk about predictive analytics because this is more advanced. What can we do with this AEC data for AI machine learning? I gave some examples here on the slide. For example, if you want to predict the issues-- because I believe a lot of the audience who are using issues model in ACC-- you want to know when you're putting, let's say, like 100 issues on a project, you want to know which issue to solve first.

So this really relies on the issues priority level. Most of the cases, the superintendent on the construction side probably based on his or her experience, subjectively choose the issue to solve because he thinks or she thinks this is more important. But in order to put in a more objective way or using an algorithm, so we're using the different parameters, like business unit, issue type root cause if it has impact on schedule or cost, which company the issue is from, and what's the trade the issue is liaising?

Based on these eight parameters, we're predicting the label, which is the priority of the issue. So in this way, systematically, it's going to tell you, this issue is at a high priority. That issue is a lower priority. So you could objectively choose the issue to solve first based on the priority level. And the other use cases like time series analytics, this is more associated with, for example, if you want to predict the tokens consumption in the next year, if you want to predict the labors, these in the next year for your factory, et cetera, it is based on the timeline.

There's more involved with the statistics like the exponential models like triple extension model, double exponential model, or robust remote model, this lies in this area for the time series analytics. Another example is NLP and LLM. It stands for natural language processing and large language model. For example, you put a lot of descriptions of the issues on the construction side. Based on those descriptions, I want to see which issue contains more risk, right?

This is another perspective to analyze the issues priority level by using the NLP and LLM modules, and the description, the text words you put in there associated with the issues' descriptions. There are a lot of different possibilities speaking of AI and ML's application in the AEC industry. If you're interested in those areas, we could talk more. And you could use our consulting services to tackle those areas. That really wraps up the technology part that are associated in this presentation. I will pass it over to Tommy for the conclusion.

TOMOTOSHI JO: OK. Thank you, Liang. That's all for the session. And next, we will go to the Q&A.

LIANG GONG: Thanks, everyone.

TOMOTOSHI JO: Thank you.