Everyone, welcome to Herb's hot takes. I'm Tyler kern, and I'm here alongside the man, the myth, the legend, the man himself, Mr. Herb Billings, VP of technology strategy at Datascan. Herb, how are you? I'm doing great today. How are you, Tyler? I am good. I'm ready to put my math hat on as we discuss record accuracy formula. How are you feeling about this? I'm feeling like this is going to be a challengewithout a chalkboard. I wish we had one these that I can pull down from the ceiling or something like that, but we do not have a chalkboard, so we're going to have to get theoretical with everybody. And also, it's worth pointing out, you have a blog post on this. And so if at any point the math information gets a little dense for people and they need to see it, they can also go and read that blog post where you have a lot of this spelled out there as well. Sure check out Datascan's blog, and that'll be a good source. That will be a good source. So check that out. So but today we are covering record accuracy formulas and in asking what record accuracy formulas people should be using. But first, we should really discuss the various options that are out there and discuss what kinds of formulas there are and how they differ. Sure, the RFID world loves to talk about the percentage of SKUs that have accurate counts, and they can do that because they're capable of counting every week and they can get things very accurate and not take the time between counts that caused the degradation of inventory, record accuracy to get too serious. So they're hitting 98% of the SKUs with accurate counts. And when you're talking about that level of accuracy, that's the kind of formula that makes sense for them. There's the some absolute variance or some absolute difference formula. The sad formula. Well, that's one way to call it. The absolute difference is going to be the count. It's actually the sum of the absolute variances unit variances divided by the grand total units in the store. That's better for a barcoded retailer because it does give you a little bit more information about the distance from accurate. You are of skews inaccurate, doesn't give you an idea of how far off you are. So that's one of the benefits some absolute difference. The third one that we've just come across by Netapp, it was a white paper by that company that makes a whole lot of sense. And correct some of the issues with the some absolute difference method that's short smap that's the short name is smap. It is symmetric, mean, absolute percentage error. And that is the last time I will say that we will call it smap for the rest of this, this, this episode. That one is very interesting because it values, it assigns a value to each variance and a one unit variance can really hurt you if you're thinking of 0 versus 1. But it doesn't hurt you so much if you're thinking 99 versus 100. If you go back to the said the sum absolute difference method, that one hurt you the same. And if you go back to the percentage of inaccurate SKUs method, then that also hurts you the same there. So it's, it's a very good improvement over the other two. This method is now our favorite method to discuss, because baked into it is the concept of the omnichannel digital risk issues. Right and we've talked in depth about omnichannel and that sort of thing in the last couple of episodes. If people haven't checked out, that conversation's worth going and diving into. Why omnichannel and why this approach kind of makes so much of a difference when we have these conversations about inventory, record accuracy? I think now that we're through the acronyms, it should be a lot easier to discuss. I think you're absolutely right. And so I suppose when I think about this, I always wonder if a retailer were to come to you and say which formula is the best? Is there a one size fits all that you think is best across the board? Or do you need to take some factors into account and ask, OK, well, are you SKU or barcode or, you know, are there other things that you need to know about a particular. A retailer, before you can tell them which one would be best for them, at my age, one size fits all is never a good option. It's a myth. I've not seen that in this case. That's also true. The RFID world does need the percentage of accurate SKUs or the percentage of inaccurate SKUs. That's the best for them given their environment and issues associated with RFID. Sure on the barcode side we do like the formula best. If you're going to have just one, it will give you the inventory record accuracy score in a way that does not value the same variants the same way when one hurts you more than the other. Yeah, that 99 versus 100 example makes a lot of sense as compared to 0 versus one, because if you have 90 you know, if you have hundreds or if you have 90 nine, at the end of the day, you're going to have something in stock for somebody if they go and buy it 0 versus 1 whole different ballgame. And as we discussed on the previous episode, then you suffer the effects of poor inventory accuracy much more. You know, you feel those effects much more when it's 0 versus 1 or one versus 0 as opposed to 99 versus 100. Yes the sum of absolute difference method, which is what the most common one we see among retailers, it's very important, but it's just not the best. If you're looking for just one in many times we recommend having both or many times the retailers there, they're the people that are interested in this, the buying organization or the different organizations and the retailer. They will be interested in both for different reasons. Interesting interesting. OK, so I'll be honest that part of the reason I am I am a host is because I was told there would be very little math and this is quite math heavy for me here. And so that's just where I'm coming from. So when it comes to these formulas, tell me about implementation and how easy they are to implement in a retail setting. If somebody is listening and they think I am metaverse. Well, I will say that programming is my background for many, many years and the actual formula itself is not terribly complicated. So these are relatively easy to calculate. But as we're finding in the data science world, clean data is the most important thing and cleaning data is it is 85% of a data scientist job, in this case, of accurate. Skewes is not affected by poor data typically. So it's not affected by, let's say, your nonrevenue SKUs that you might carry in your ERP system, but not actually count in your inventory. So those will be inaccurate, but you're going to have one SKU that's inaccurate or 10 SKUs that are inaccurate. Out of the 10,000 you have in your store, the sum of absolute differences is very sensitive to that. You might have 5,000 units in your ERP for the gift card and you're not going to count any of them because that doesn't make any sense at all. You'll have a variance of 5,000 units in that case and your inventory record accuracy will definitely suffer as a result. Those need to be scrubbed out of the calculation. And the first time you do this in my history as a programmer, the first time you do it is fine. You might hard code or put this in a table somewhere and say we don't want to calculate these, but if you don't have a specific, these are not counted or these should not be part of the calculation flag, then that has to be constantly maintained. And if something changes in the future, askew gets added that shouldn't be counted. It would have a significant impact on your inventory record accuracy. And since this is going forward, going to be a KPI for many stores. You know, people take that stuff seriously. And when it does change, especially to the negative, that's something that people will have to pay attention to. So data scrubbing, making sure that you keep your stuff clean is a challenge for the said method. The method is less sensitive to that. So you don't really need to keep your data is clean for that. Really interesting, especially getting your perspective as a developer, I think is particularly fascinating. So I'm wondering now is as we look towards the future, other trends or developments that you're seeing that are coming when it comes to inventory, record accuracy, what do you see on the horizon? Well, it's becoming more important for retailers as a result of the channel, the rise of the omnichannel with. It was before the pandemic, but then the pandemic Press the fast forward button, as we've talked about before, certainly did people who have never bought online and said they never were forced into doing that and are starting to like that experience, at least for some items. Inventory, record accuracy is very important for retailers. The trends we're seeing or they're starting to use that as a KPI for stores. It does a very good job of giving an overall picture of whether the stores following their processes and procedures appropriately around inventory management. There are other related data dives that you can do from your inventory record accuracy into the variance side of things. You can see where things are located. So your items that are in the back room, not on the sales floor, while by itself not related to inventory record accuracy, it's definitely related to location accuracy. And you want this where people can see it has inventory record accuracy risen up the food chain in terms of who is concerned with it, it seems like it used to be a thing before channel, before buying online and that sort of thing, that this was something that may have just been an annoyance for people who worked in the store on a daily basis who had to say, oh, no, we don't have that or let me go check. But now it seems like this is an issue that has climbed the food chain to the point where now higher and higher UPS are increasingly more concerned about inventory, record accuracy. So knowing that you and I have talked before about how life looked prior to the channel, prior to the need for very accurate inventory records, it was an annoyance at that time. It was something that retailers counted when the stock levels were lowest. That would be the lowest amount of labor that they would need to count their stock. They could get the reporting, their shrink calculated. Everything was done as cheaply as possible. Today, it is a sales strategy. We've talked in the past about how you can increase your sales between 4 and eight percent, according to one very large and very well done academic study by counting more often. So when you talk about increasing sales, that C-suite level and these folks are very interested in how to do that, especially for the low cost of counting your stock more often at relatively low cost. So it definitely has changed the schedules there, counting when it's much more advantageous to have accurate inventory for their channel, big selling seasons. They are changing their frequency as well. They're counting more often. They're counting the most important items, more often resulting in increased sales and increased visibility in the organization. that's a really, really interesting and you're right, any time we're talking sales numbers, that C-suite level, and that's something that those guys are very, very concerned about. So far. We reached the end of this particular episode. Let's tie a bow on it. What kind of conclusions or summaries do you want to leave the audience with today as we wrap up this episode talking about formulas? Well, check out our blog posts on the subjects. They will give you the positives, the negatives of each one of these methods, also the details of how to calculate it and what they're good for. Inventory record accuracy is very important for all retailers, but extremely important for omnichannel retailers. It is the foundation of their success. All omnichannel orders start with product availability. Product availability is very dependent upon inventory. Record accuracy increase your sales by counting more often. And you know, that's what retail is all about. Absolutely true. Symmetric mean, absolute percentage error. Wow, I got it smap there you go. Go check out the blog post for more and stay tuned on data. Scan for more episodes of herbes hot topics. Also drop us a line if you want us to cover a topic. If you're a retailer out there and you have a question for her, maybe drop us a comment or a question and we could do a future episode over something that the audience suggests. I think that might be fun. But everyone. Stay tuned for more episodes of herbes that takes her Billings. Thanks so much for joining me for another episode. Thank you very much. Absolutely And everyone out there, thank you for joining us for this latest episode of herbs. It takes, like I mentioned, stay tuned for more episodes of the show. But just make sure to bookmark data, scan dotcom so you can get herbes blog posts. You can get the latest episodes of her outtakes takes and stay tuned for more, but until next time. Herb Billings, I'm Tyler Curran. We'll talk to you again soon.
