Welcome to What the Duck?! a podcast with real experts talking about direct spend challenges and experiences. And now, here’s your host, Source Day’s very own manufacturing Maven, Sarah Scudder. All righty, welcome to our weekly What the Duck?! Another Ducking Digest show. We are doing a multi-part series on the show this month, talking about technology adoption and Tech implementations for small and midsize manufacturers. Last week, we went through three of the six best practice guidelines for success for an implementation, and this week, we are going to discuss the last three steps, so steps four through six.

Lindsay has what I consider to be a really good example of utilizing drones to improve warehouse management to talk through and illustrate some of the points. So, Lindsay, I will let you dive into the last three steps in our list.

So, next three steps. Last week, we talked about an explicit problem statement, second having a living document of record of what it is we’re doing, and thirdly, engaging a team, not trying to do a solo effort. You know it’s going to take a village.

For today, the back three are identified, detail, and celebrate interim milestones along the way in our project. And then, hold on to the discipline, hold on to change management, hold on to program management, and hold on to reporting. And then, six, take the extra time to sweat the details. And the example we’ve got today is a good one that shows how things get complicated. The simplistic idea: “Yeah, let’s use drones to confirm something is in a bin location or in a pallet.” So, instead of broad-brushing that as an explicit problem statement, the automated verification of all SKUs and part number barcodes via drone, you know, before we launch the program, we want to take it down to a couple of levels.

Surprisingly to me, the first thing we have to do around the viability of this is confirm we can fly the drone in the warehouse. So, just as every Supply Chain’s unique, every Warehouse is unique. You know, unique lighting, unique ambient light, unique artificial light, unique ceiling. The ceiling uniform.

If I set my drone to stay below 21 feet, am I good, or is there, you know, what’s the schedule for high lifters or forklifts going up and down aisles? Are all the aisles the same? So, the first milestone that we want to drill into and successfully fly my drone down one aisle, and then the next step, okay, I can do it with a serpentine, and then can I identify, can I avoid collision by either seeing something ahead of me and reversing or stopping? So, that would be step one for drone flight viability. Excuse me. Then the second thing, a low-hanging fruit, an easy win, a drone alert for an obstructed aisle, and that can be, again, that can be anything, and the output is a warehouse housekeeping report, if you will, that says, ‘Hey, there’s something in aisle 17.’ It could be Lindsay and Sarah going around looking at boxes. It could be a carton that’s fallen off a pallet. It could be a critter. It could be an AMR or a forklift that’s stalled or being parked there. But the point is we’re going to be able to identify whether or not the aisle is clear as part of our fundamental housekeeping. The question then is flight time. How long is it going to take for my drone to do that? If I can do an aisle in 40 seconds and I’ve got 20 aisles, is 14 minutes flight time acceptable, and if so, when might I do it? My favorite is to do it between shift changes. So if the morning shift kicks in at six o’clock, do a housekeeping flight at 5:30 so that the team, when they show up, have this report. So another interim easy win, building confidence, generating familiarity. Next fun one is to create an alert from the drone on an empty location. Now, this one we underscore the previous criteria on how much time is this going to take because now we have to be able to scan if a pallet B has something in it. I don’t care what it is, is there something or is there not something. So I have to train a drone, I have to train a camera, I have to train a camera to recognize an empty B, and that probably takes about a thousand images. Here’s what an empty B looks like, and we’ll build this composite. So that’s going to take a while. Now, that’s going to require AI support to code it, and then the question of the physical mechanics of my flight time. Now, instead of being, what did we say, 40 seconds for an aisle, now if I’m 5 seconds per pallet location, and I’ve got 600 pallets, you know, is it going to take me, and how can we fit it into the schedule of the work? What we’re trying to do is augment with AI but at the same time fit into the culture of our warehouse and the constraints. The important thing for today’s audience, Sarah, is that it cannot answer the question, ‘Is this going to work?’ The point is supply chain, warehouse operations, supply chain, there’s a subject matter expert who owns the context and is able to say, ‘Yes, I can plug my 15 minutes in here, yes, I can live with an hour for an empty bay verification,’ and then we start to get, you know, we’re just building and building, right? The next layer might be to correlate my drone camera information with my ERP data. So, to begin with, don’t worry about where the ERP data came from; we’ll get to that. Does ERP agree? Now, some folks might be scratching their heads, saying, ‘I’ve got 99.99% cycle count accuracy. I don’t need this.’ And certainly, if that’s the case, then we don’t want to invest in solving a problem that doesn’t exist. However, if we’re in a small company that’s prone to staffing changes and sometimes discipline moves in and out, depending on the team we’ve got or the number of employees or our staffing level, then maybe it’s something we want to have in the back pocket. If things change, we know we can do this. The ‘got you’ from ‘shouldn’t have to do this,’ the ERP already tells you this information. You know, this is a housekeeping issue. Yep, guilty as charged.

Thank you for your input. This is what I want. It’s an error. You’re exactly right, but it’s not an employee performance issue. As far as it is concerned, this is an error condition, and I need you to give me an error report. So, that’s just going to be part of the spec. It gets fun because it goes in two different directions. One, the ERP says there’s nothing there, and the drone camera says, “Yep, there’s something there.” So that triggers our subject matter experts, our warehouse team, our housekeeping team to let’s go take a look. Let’s find out what’s happening.

The second one is that the Drone says there’s nothing there. Let’s go find out. Let’s do a cycle count reconciliation on that. What we can do once we’ve got that in place is we can publish a scorecard dashboard report that says, “Here’s my open bay availability, and here’s my capacity.” We could do that with our ERP, but now we can do it with the underscore of the physical feedback. The cameras, in an ideal world, are just providing physical feedback of what we believe to be the truth.

Then we get really exciting. We can use the camera to read a barcode. Okay, it’s very easy to do it in the lab, tricky to do in a warehouse, especially if we’re challenged by pallets not being oriented the same way, barcode labels not being in the same location, barcodes being torn or wrapped with one or two or three or six layers of stretch film. Makes it hard. Cameras don’t like stretch film over barcode labels, so we might have to have an interim report. Can I read how many? Can I train the camera to read the barcode? Can I train the camera to find the barcode on the pallet, recognize it? Again, another thousand images, perhaps, to go and locate the barcode on a pallet.

Of my 18,800 pallet bays, how many of the 1,600 can I successfully read, and then give me a report that says, from a housekeeping standards, “Here’s the ‘could not read’ report.” That’s just the example for the camera. For the Drone, now we have to go through all that for the network, for the data stack, for our IT data merge, for our employee training, for KPIs, for reporting, for workspace. What we’re doing is we’re building out the cliché of knowing what success looks like.

And so, but as well as that, we’re also getting a grasp of the complexity we’re going to get ourselves into. We have to agree up front how we measure if we’re successful. How do we avoid something not sticking? Things won’t stick. It’s not if they won’t stick; they will not stick. So what do we do about that?

That was this morning’s drill into the milestones, program management, change management, working the details. That’s all I’ve got for this morning. So with that, Lindsay and I will be back next Monday to talk about something that is relevant to those working for small and midsize manufacturers. Have a wonderful Monday.