Lost in the Labyrinth

Kevin Van Leer
Nov 8 · 6 min read
Photo by Maksym Kaharlytskyi on Unsplash

I spent the last five years of my career at a startup that recently closed up shop, basically overnight. We were in the middle of three promising pilots with three of the largest 3PLs in the world. How could this have happened? Looking back, it’s pretty clear that we had been dead for over two years. Living on through tantalizing promises of big contracts that were just over the horizon.

Getting started

In 2016 we partnered with a large company to provide a visualization layer to their new IoT product. They were marketing an indoor location tracking technology using Wi-Fi triangulation and needed a visualization layer that could display the data they were collecting along with other data generated by customers. Together, we were trying to win business with a large third-party logistics provider.

It sounded like the perfect relationship. Prior to this engagement, we had been developing a SaaS platform for spatial and temporal data visualization. Our mission was to ingest any data with spatial coordinates and a timestamp, then display it on a map, giving users the ability to interactively watch the data as it occurred. We were looking for markets for this technology. It sounded like the perfect fit. And it was just in time. We needed to start onboarding customers and generating revenue.

The customer was looking to pilot a connected warehouse technology. They wanted to create a digital twin of their logistics operations using activity from the warehouse management system (WMS), usage data from material handling equipment, and location tracking from Wi-Fi scanners carried by workers. Again, it sounded like a great fit. Our platform was designed to join and visualize data from disparate sources.

The pilot hinged on the viability of the Wi-Fi location tracking technology. Use cases revolved around decreasing distance travelled and safety. Detailed positioning data (the customer specified 1 meter resolution) for equipment and workers was touted to be the key to gaining insight into these inefficiencies. We’d already been through the proof of concept phase before I joined the project. There were accuracy issues, but everyone was confident they would be resolved.

Nuts-and-bolts

We pushed on with the project, kicking off the pilot in October 2016. The biggest technical challenge was building links between the location data and the other two sources. In theory it was easy. Locations were tracked by MAC address, each MAC address could be associated with a scanner ID. Scanner IDs were recorded in the WMS transactions along with employee IDs. Using that path, we could correlate location data with WMS activity. From the other side, users activated forklifts and pallet trucks with keycards. The keycard registered a user’s name in the system when a piece of equipment was activated or deactivated. User names could be linked to employee IDs in the WMS. In theory we could connect all three data sources.

As one might imagine there were a lot of issues with this plan. Lookup tables had to be kept up to date, and otherwise wise trivial details like typos in the spelling of employee names in MHE usage data became quality issues. As significant as these issues were, they only impacted a small minority of use cases.

Another significant challenge for us was representing the WMS activity in space. Since we were also visualizing location data, we needed a scale representation of the warehouse. The CAD model was the logical source from which to draw this information. It had both the overall dimensions of the warehouse and the location, size, and shape of the rackings, bulk areas, staging areas, and loading docks. All the location information we needed to present the data. The next step was identifying all those locations in such a way that WMS data records could be associated with locations. That turned out to be fairly labor intensive, but it was the pilot, we’d find a way to streamline the process.

By January 2017 we had automated the WMS data feed. We were still in the process of setting up a site-to-site VPN for the location data and some of the infrastructure was being upgraded in hopes of improving location accuracy. Things stalled out over the next few months while the customer and vendor tried to figure out what to do about location accuracy. Since the technology was embedded in Wi-Fi access points, upgrades required infrastructure level changes. It was a slow process.

Pivot

In the meantime, I had started to realize that the WMS visualization we’d built could stand on its own as an analysis tool to help plan warehouse design and improve warehouse efficiency. This was confirmed at an onsite visit with the customer in August 2017. We received mixed feedback from folks at the site. They were still anticipating a solution to the location tracking accuracy issues. When the WMS planner saw what we had built, he was immediately immersed in the visualization. He was able to recognize events that had occurred in the days prior and operational inefficiencies born out by patterns in the data visualization. It was the validation I’d been looking for.

Unfortunately, that basically meant starting over with the customer. We hoped we could transition the focus of the pilots to WMS operational monitor, but the project had been sold on the promise decreasing distance travelled through location tracking, and there were already WMS analysis tools in product, so starting over took a long time. Before we could bring them around, we did location based pilots at three other sites, and at one, used a second Wi-Fi positioning technology. All of them led to disappointment and were eventually cancelled. It wasn’t until early 2019 that customers started coming around to the idea of a WMS operational monitor built on our platform.

We had found the problem that needed solving, but we weren’t radical enough in the way we repositioned our product. We were still marketing a high fidelity, real-time, always connected, operational monitor. It’s a great vision for the future, but it doesn’t help real people in real warehouses solve the problems they face today. We didn’t provide a path for our users, or our customer, to take their operations from where they are right now, to that ideal end state.

By then it was much too late anyway, we’d fallen short with investors, and we couldn’t get a purchase order until the customer had proven a positive return on investment. Even if we had, the solution we’d developed wasn’t easily scaled. The rate of adoption was so slow that we wouldn’t have been able to generate enough revenue.

Hindsight

With the perspective I have now, I can see that the critical juncture was the proof-of-concept. I think we saw the signs at the time, but we really needed the business and no one was willing to ask the hard questions. The accuracy issues with the location tracking system were not kinks that needed to be worked out. They were fundamental issues between the technology and the warehouse. Physics got in the way. Before it was all over, the customer and vendor would spend over two years trying to increase accuracy to the point of viability. We didn’t have two years.

We were trapped by the requirements of an unviable solution. If we had started out building a solution focused on WMS, the need for a high fidelity spatial model of the warehouse would have evaporated, and the requirement for automated live data feeds would have been deprioritized if not dropped altogether. Shedding these restrictions would have freed us to develop a solution our users could use immediately, on their own, without our assistance.

That would have changed things dramatically. Instead of engaging with IT departments coordinating integrations, we could have engaged directly with users, allowing them to demonstrate value from their first interaction with the product. It would have unbound our team from high touch requirements gathering and use case generation. We could have automated our sales pipeline, allowing for a totally self-service experience.

Whether or not any of these changes would have improved the company’s future is anyone’s guess. We were on a timeline that we could not sustain, and our success was coupled with an unproven technology. Then, when we became suspicious that the path we were on was leading us in circles, we didn’t follow our instincts. Ultimately, that inability to adapt is what led to our dissolution.

Grasp Data Inc.

http://grasp-data.com

Kevin Van Leer

Written by

Kevin is a co-founder and Director of Product Development at Grasp Data Inc.

Grasp Data Inc.

http://grasp-data.com

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade