deal architect

Uptake Verticals - Part 2 - Dealers of Industrial Equipment

This continues our conversation about verticals at Uptake. Yesterday Lisa Dyer talked about their "asset and maintenance engines" and how they are adapted to several industrial verticals. Below we cover the discussion with Joe Rosing, Director of Product Marketing about functionality for Dealers of Industrial Equipment.

The Functionality

"I've been with Uptake for about a year now.  Previously, I had a little over 12 years with Rockwell Automation in a combination of manufacturing roles as well as product management in the control systems business.  Here at Uptake, I have focused primarily on solutions marketing with our heavy industries, specifically our dealer line of products that serve dealerships with construction, mining, and ag assets, as well as mining operators and then manufacturing as well.

 Most dealerships for heavy equipment are split primarily between a sales function and a service function.  Within those functions, then you'll have individual divisions.  For example, earthmoving or a division for agriculture or a division for rental, the motions of that business and the customer base are slightly different.  But what comes across all of those divisions, really, is the sales and the service workflow.

SalesLink is what we call our industrial CRM.  At a certain level, a CRM is a CRM.  But we've built it specifically for a dealership business.  So, it allows you to click into things like, "Show me all of the pieces of equipment that a customer has.  Show me all of the service notices or the warranty information for each of those pieces of equipment."  It gives you that level of details that, as a sales function within a dealership, you're going to want and you're going to want that pre-configured for you versus going with a more generic CRM that is maybe purposed for a broader set of industries.

Then in a similar way, what you have here with this iQuote and iMacs, these are tools to help you do intelligent quoting.  The reason why this is specific to the dealer business is because it allows the dealerships to look up, in real-time, the current price list with the OEM to make sure that they're getting the best available price at that time.  Otherwise, what you have to have is either people go online to the Caterpillar website or you might have to build other APIs into other systems.  We have all that purpose-built, so it keeps everything in sync with the latest and greatest information that's available from the OEM or from any type of other promotional activity that the dealer might be doing.

Then COTI is an application that helps manage the workflow once an order is taken through invoice.  If you think about what's involved in buying a bulldozer, for example, while the Tonka trucks all look the same, in reality, they're all configured differently.  A lot of times these machines are basically configured to order from standard components.  Those components might be coming from different places.  Some of them might be in stock at the dealership.  Some need to come from, basically, a warehouse through the OEM.  Some might need to be built.  They might be custom or just lower volume and they need to put them on order a full lead time from Caterpillar.

There's a lot of touchpoints and a lot of people involved.  Actually, once the order is taken, getting it to the point where all of those components arrive at the dealership.  Once all the components are there, getting it assembled into the final product as the customer would expect it.  Then managing the delivery process to the end-user and, ultimately, generating that invoice.  Whereas, in the past, that whole process might happen between a number of different systems or on paper, sometimes on whiteboards, what COTI does is it helps create, basically, one central place that manages that end-to-end workflow for the dealership.  Those are primarily the applications for the sales workflow.

Then for the service workflow, we integrate with some of the same components like this industrial CRM.  We only want you to have one CRM as a dealership.  We don't want you to have to think that you need multiple CRMs for different divisions or different workflows.  But, really, we start one step in advance of this.  This is where we're starting to use the asset data as a way to generate new leads for the dealership based on service opportunities.

We were running it initially with the condition monitoring portal, which connects to the asset data that's available on all of the equipment that you have basically set up with the system.  That allows you to get access to the different machine alerts that are coming off of the IoT devices on the machines.  Then there's a condition monitoring analyst that goes through those alerts, filters out the ones that need to be actioned, and then create a case.

That case is turned into a sales lead for service that says there are five fault alerts on the engine that a technician should go look at.  Whether this piece of equipment is under warranty, under a CSA, or just a straightaway service sale opportunity, it goes into the CRM as a lead for a parts and service sales rep to call on that customer and say, "Hey, look.  The data off of your machine is saying a service needs to be done.  Let's talk about how we want to handle this."  That's why it's integrated into our CRM.

Then, from there, we have, again, some quoting tools that are specific to that use case.  In particular, parts and service quoter.  We want the dealership to have access to information like, again, is it on a CSA?  Is it covered under warranty?  Do we want to charge time and labor or time and materials?  Do we want to charge just labor only or materials only?  You have that flexibility built into this parts and service quoter that is a little bit more intelligent and prepopulated than you would just out of a standard quoting tool.

Then once the customer says, "Yes, I want to do that work," it gets saved as a work order in ServiceLink which is really a service work order management system.  You can schedule your technicians in ServiceLink.  You can have visibility to making sure that you have the right technician assigned to the right job or, if they're going to be in a certain geography at a certain point in time, you can have them based in that geography for a week, for example, and look at the four or five service opportunities in that geography and get those assigned to them.

All of the work order detail is saved and passed through the platform here.  From the point in time that a condition monitoring analyst identified those fault codes, that service technician, actually, onsite at the machine, can see what fault codes were identified, any notes that that analyst put in place.  All of that information is available to them.  That's the advantage, again, of just having a platform that serves the different points along that workflow, both for sales and for service.

The Value Proposition

When we talk about automated case recommendation, this is the AI package that we had built inside of Asset IO.  This is a new functionality that we just released over the last month that starts to automate some of the process that we established with the condition monitoring portal.  It starts to automate what the condition monitoring analyst is doing by looking at the history of the data, whether that be machine data, work order data, case creation history where we can identify, for example, if we see these same four fault codes, those four fault codes typically end up in a case that gets actioned by the dealer.

We take that as a key input into this recommendation so that the software now can automatically say, "Any time I see these four fault codes, I'm automatically going to create a case and automatically populate a description of what could be wrong and then add a prescription for what needs to be done to fix it.”

With our insights engine now from Asset IO, what we're finding is 89% of the insights end up being reliable cases.  Now, think about what that means for an analyst.  They can look at 50,000, let's say, machine alerts and put one percent of those in a case, or they can look at hundreds of insights and put almost 90% of them in a case

The vast majority of our data science team has been working with this dealer business now for about four years because Caterpillar was one of the initial customers of Uptake. We also send them out regularly to go visit dealers.  In particular, with developing these additional data science models, there is a lot of interaction, either by phone or in person, at the dealerships to gain an understanding.

We still rely on customers for deeper level subject matter expertise.  But I can tell you, to put this together, there were about 30 new data science models that were developed and a lot of those were developed with the subject matter expertise of our data scientists that have just developed that expertise over the last couple of years.

I would add that I think the other place where we have expertise isn't just in the asset, but in the process itself.  No one really talks about assets in manufacturing.  You always talk about process.  Everything is at the process level.

As we get into automotive cases, for example, or metal fabrication, you're not just looking at an individual stamping press.  You're looking at the end-to-end stamping process.  We developed expertise in how material moves through that process, the robotics that are involved, the control systems that are involved, the data that's available.

From an expertise standpoint, really, what you're looking for is, as you talk with an operator who has been doing stamping on an individual line for 30 years, they almost have this machine whisperer talent, right?  They can just tell by the noise a machine is making or by the vibration of something on the line that there's something wrong.  What we're trying to do is basically marry that intuition with data-driven precision.  Beyond just having knowledge or expertise around the asset, I think we've done a lot of work the last couple of years to develop this expertise around individual processes that help us accelerate implementation and time to value for customers.

The other point is that our customers, some more so than others, are thinking about their workforce, their aging workforce.  What do we do with all the intuition and the knowledge that walks out the door when these folks retire?  This is just another way of codifying that deep, deep knowledge and making sure that you have that as your workforce evolves.  That's one of the concerns that we have for customers as they think about their future is the workforce part of it.  In other words, it's assets, it's process, it's people.

That is a huge concern in this dealer space because the availability of technicians that can actually do the work to fix these machines has just been dwindling.  Every dealer we talk to talks about how hard it is to find good technicians, to retain the talent that they have as some of their best technicians start to retire.  How they're trying to bring new technicians up to speed.  All the more information that we can provide to them through this condition monitoring portal of, "Hey, these are the alerts.  This is the description of what's typically wrong when we see these alerts.  This is the prescription of what you should do to solve those problems."  All of that helps our dealers just be more effective with the technicians that they have and improve that technician efficiency also such that when we indicate that they need to go visit a customer to do a job, it is a job that needs to be done.  They know everything that needs to be done on that machine, so they don't have to go back out two weeks later to do some other minor repair, but they can get it all done in one shot, improve that first-time fix ratio, and then move on to the next customer.  Improving that technician efficiency is a key component of the value that we drive in the dealer business in particular.

Uptake Verticals - Part 1 - Industry Asset Engines

I had a chance to talk to two executives at Uptake about vertical markets they focus on, under their tag line of "Our technology has seen millions of machine failures so you don't have to". First are the notes from the conversation with Lisa Dyer, till recently their Director of Product, about their "asset and maintenance engines" and how they are adapted to several heavy industrial verticals. Tomorrow, we will  cover the discussion with Joe Rosing, Director of Product Marketing about functionality for Dealers of Industrial Equipment.

The AI engines

I've been at Uptake for four years now and since the beginning, we’ve always been really focused on the industrial space, industries with heavy industrial assets, with the goal of helping them to increase productivity, reduce downtime, financially optimize their maintenance strategies and, ultimately, help them to improve their own customer experience.

So how do we do this?  We have an APM application called Asset IO. Broadly, it offers a full suite of what I'll call blended maintenance strategies.  Meaning, a company - given their journey in transforming how they manage their assets - may be at different places in terms of their maturity and thus, appetite for different strategies.  For example, not everyone is ready to go all the way to predictive.  The blended maintenance strategies and offering that we have helps customers understand what they can do now, given where they are from a maturity standpoint, and gives them a path forward to advance their strategies.

Let’s talk about the ground work we’ve done with the Uptake Platform that ultimately supports our applications and provides our customers with an end-to-end solution.  The platform uses a series of AI methodologies to ingest and normalize data as well as predict failures and maintenance needs.  We have over 150 unique algorithms that serve all different types of assets and industries.  To create new models, we use our AI engines, which serve as a starting point for common use cases.

For organizations performing reactive and/or time-based maintenance, for example a time-based maintenance plan from an OEM, we can help them create quick wins for their organization by reducing over-maintenance of assets.  With our data integrity engine we clean up work order data by identifying missing or inaccurate data, using unsupervised, supervised machine learning techniques and text mining.  We use AI to automatically make improvements for missing values or records to create asset and failure mode tags for each operator action so we can standardize across the plant and the fleet regardless if a customer is using different EAM systems.  This mitigates data quality issues so customers can see true O&M cost analysis.  Then with our planned maintenance optimization function, we can perform what-if scenarios to find the optimized maintenance strategy.  Our asset strategy library, which is a catalog that contains something like over eight hundred asset types that are supported by over fifty-eight thousand failure modes that are found in over a hundred and seventy-eight thousand conditions, feds into the PM optimization function to allow what if scenarios to be conducted whether customers have connected assets or not. This helps them balance over and under maintaining assets.  One customer was able to reduce their maintenance costs by over a hundred thousand dollars a year by optimizing their maintenance strategy for 25 voltage motors across one plant.

We also can help organizations with other maintenance strategies like condition monitoring and predictive analytics.  We use our MLAD engine, that’s Machine Learning Anomaly Detection, for condition monitoring which detects abnormal behavior in asset data.  It determines the high/low tolerances for an entire fleet of assets.  This was helpful for one of our wind customers.  After just two weeks of monitoring the entire wind farm, we were able to determine one turbine was performing abnormally compared to the others. We alerted our customer to a potential failure and when they checked the turbine to determine the main bearing was about to fail. The repair cost them five thousand dollars verses the two hundred and fifty thousand it would have cost them had the bearing failed.

Additionally, the ‘normal range’ expectation adjusts as operating circumstances & environments change.  For example, a customer won’t see an alert when a spike in engine temperature occurs because of an external cause like increase in payload or going up a steep hill.

Our failure prediction engine ingests high volumes of  live streaming data across multiple sensor readings to detect deviations.  Instead of just understanding if an asset hits the high/low threshold, the engine correlates multiple sensor readings to detect and alert to deviations in order to provide advance notice that a failure will occur.  We caught a coolant issue because we were monitoring engine speed, engine oil temperature, and engine coolant temperature of a locomotive.  We saw that the engine coolant temperature was deviating slightly from the patterns of the engine speed and oil temperature.  Fault codes never fired because the data didn’t hit the OEM thresholds and the deviation in the data was too slight to notice by just looking at the data patterns, however our customer was alerted to the issue and able to schedule maintenance in advance of a failure occurring.

Adapting to verticals

While there are algorithms that can be used across different inquiries, most of our algorithms are built to support a particular industry.  We just use our AI Engines as a starting place to then tailor them by industry. This allows us to quickly deploy new models while also tailoring the applications.  Like in the trucking space, we have an algorithm to detect when a truck is running an inefficient DPF or diesel particulate filter. With one customer, we were able to save them over five hundred dollars a month for one truck with this model.  The fleet/trucking space is also a great example of how we tailor our applications to be industry specific not just with models, but capabilities, for example, we have driver scorecards and a mobile app for ELD, electric logging, that are needed to support our customers in the fleet industry.   

We have data scientists who are focused on specific vertical industries.  In part, what informs them and gives them the knowledge, besides the visits with customers and prospects and so forth, are regular deep dive sessions with folks who have been in the industry for 30-some years and they teach how to fix these assets, for example.  We rely on our advisors in understanding and getting down to the lowest level of how an asset or component could fail and all the ways in which it could fail, to help us find what causes it to fail and ultimately help improve the accuracy and efficacy of our data science models.

In certain industries such as transportation, the asset technology itself is evolving.  Take an on-highway truck, which is predominantly linked to a diesel engine.  Now we have CNC, natural gas-operated.  This is a whole new type of asset, a whole new process for maintaining and, again, that workforce that you may have may not be as familiar with something like a natural gas-powered or a renewable energy powered vehicle.  Our data science models that we develop around those new machines can also help with that.