Skip to content
 Attention Shopify merchants: Get an instant estimate with no commitment to book using our Flexport App in Shopify today! Get an estimate
  • English
  • Deutsch
  • 简体中文


Logistics Rewired: Avoid Stockouts & Surplus With Better Inventory Planning

Logistics Rewired: Avoid Stockouts and Surplus With Better Inventory Planning

Ted Boeglin: Hello everyone, and thank you for attending today's webinar, which is about Avoiding Stockouts and Surpluses With Better Inventory Planning. I'm Ted Boeglin, and I lead Flexport teams who helped small and medium businesses grow and profit using Flexport technology platform, our logistics infrastructure and our global trade expertise.

Today we are covering the most requested topic from our customers, which is demand forecasting and inventory planning. So why does this matter? Well, because getting it wrong means that you lose money or lose sales, and getting it right as a competitive advantage for you, it means that you can sell more product at lower costs.

Before we dive in, let's go over some housekeeping items. On your screen you will see a sidebar to the right of the main stage. If at any time you need assistance during the live webinar, please message us in the help chat located on the sidebar. You can also ask questions via the Q&A tab on the sidebar. We will be answering those questions in written form throughout the webinar, in lieu of doing live Q&A at the end. We will make a copy of the presentation slides available and we will put it in the chat. And lastly, on the menu bar located above the screen, you can find a link to a short survey for this webinar, your feedback is so important to us. We are constantly looking to improve this forum, and it's so helpful when you give us feedback about the topics that we cover.

A brief legal note, keep in mind that all the information we provide in this session is based on the current situation. It might not be customized to your specific business requirements. And we always recommend that you reach out to Flexport, talk to an expert to discuss your particular situation. But with that behind us, let's now meet our guests who are going to be with us today. First, I'd like to introduce Alex Kopco, who is the Co Founder and COO of Forum Brands. Alex, welcome, thank you for joining us. And can you tell us a little bit about Forum Brands?

Alex Kopco: Yeah, thanks so much Ted. I'm really really excited to be here. As you mentioned, my name is Alex. I'm one of the Co-Founder Forum Brands and the COO. What we do at the company is we acquire small digitally native brands and we scale them globally to be big digitally native brands. Additionally, we recently launched a an inventory and demand planning tool called Crystal, which I'm really excited to talk about today in the context of how to make smarter planning decisions.

Ted Boeglin: Alex, I can't wait to dig into that. I was excited to see your launch of the crystal tool recently. I'd also like to welcome Pratap Ranade who is the Co-Founder and CEO of Arena. Pratap can you tell us what Arena does?

Pratap Ranade: Yeah, thanks Ted. It's great to be here. Yeah, Arena, we founded the company to basically make enterprise decisions autonomous. And the general premise is we've seen a huge amount of progress in AI, especially when you look at game playing and robotics. But then if we look at enterprises, a lot of decisions are still in a paradigm of like decision assistance and advanced analytics, and so we're founded to basically bridge that gap. We're super excited about today, we started with most of our work actually on the demand side. And as everyone on this call is no stranger, demand and supply are very deeply connected. So through that work we've been pulled into supply.

Ted Boeglin: Pratap, no matter where you start, supply chain will find you. So the last two years have got us that. Okay, so let's dive into some of the questions. Alex, I actually want to start with you. You have been so hands on with so many different brands over the last few years.

Just-in-Time vs. Just-in-Case Operations

And I think some of the buzz that's out there is that we see brands starting to question the wisdom of the just in time inventory model that was so popularized in the in the 2000s. In some argue that we're moving more towards a just in case inventory model. This is a question that we're getting a lot. How would you advise brands decide between those two models of just in time versus just in case?

Alex Kopco: Yeah, it's a really good question. And I think the term just in time itself is a little bit misunderstood. It's sort of just in time, to me and makes me think of this, you know, almost custom order I, the widget isn't produced until the point of demand and, but in a different context just in time, for example, in the context of a retailer is really about making sure that when you walk into a store, you know a shelf doesn't have 500 widgets, maybe only as five widgets and so, it's really more about ongoing replenishment making sure that there are always enough widgets to fill the shelf. But that, you know, you don't have to hold as many widgets in the in the back room or even in that location that state.

And so this notion of just in case versus just in time, I think are really interrelated. And it boils down to the topic that we're here to talk about, which is demand planning and demand forecasting more broadly. And look, to me just in time, and the way that we approached it Forum Brands is really about making sure that we are locating the right number of widgets products as close to our eventual customers as we possibly can. And that minimizes the period of which we might be out of stock. It also though conversely can help minimize the amount that we're holding in bulk in one particular space.

But as it relates to just in case, as we've seen, it is prescient in these times to make sure that you're holding enough inventory to weather, many of the supply chain shocks that we've seen. And so, it kind of boils down to a couple of factors that I'll talk about that, we think about, number one is working capital, if you've got the cash because you're a slightly larger business, maybe you have a specialized type of product, that doesn't change very frequently, you may want to stock up a bit more than you would normally be comfortable with, just in case. Just in case your supplier runs into a challenge, just in case, as we saw last summer, the port of LA gets backed up by 75 plus ships, and you actually can't get your goods off that boat for a month or two, which happened to us, and I'm sure many other folks.

But there's also this notion of, if your manufacturer is owned by you, if it's domestically located, if it's located to your eventual customers, maybe you don't need to have as big of a worry about this notion of just in case, just in time can still work for you, because you're in more control of your manufacturing footprint. So that's one big consideration is the working cap. The other consideration is, do you have the space and so? We have a lot of our sales on Amazon, and anyone who's sold on Amazon over the last two years has probably run up against the caps that Amazon will implement in certain brands and in certain categories. And so in many cases, you can't really hold just in case quantities of inventory because Amazon won't physically allow you to store that many goods at an FBA warehouse. So then the question becomes do you have access to? Or would it be prudent to take advantage of a 3PL or 4PL an asset light inventory storage opportunity, just in case, again, you want to weather these supply shocks. And really that boils down to your use case, it boils down to your lead times, your manufacturing footprint, your growth plans, and where your customers are as well.

And so, yeah, I think just in time, just in case, it is case by case. And I don't think that just in time as a concept is going anywhere. But what I think it will require particularly in times like these is an evolved level of thinking about balancing the risks, and the trade offs of being overstocked versus out of stock, and realizing those two the bottom line.

Ted Boeglin: Alex, it's so interesting to me, because Forum Brands has to do this across companies that you acquire a very diverse set of SKU's and products that probably have different characteristics, some with very stable demand curves, others with highly seasonal curves. How are you sort of mapping out the various SKU's that you have and choosing the right approach for the right product?

Alex Kopco: Yeah, we follow the data at the end of the day. Big data geek wouldn't be here if I wasn't. But no, I mean, really, we let customers basically tell us, with their wallets, with their behaviors, and in many cases with their phone calls and their emails to our customer support teams, what they're excited about, what they're not excited about. And so if we're starting to see traction, particularly on the more seasonal demand curves of these different products, we'll stock up in advance of that and we'll look for opportunities even if they're a little bit more expensive in the short term, to delight customers who are telling us that they want these products.

Those that are more stable for which we can generate a much higher probability, a much higher likelihood based model. By and large, what we're doing there is we're just holding a couple of extra weeks of inventory domestically. Again, to protect against some of these shocks, like when we saw the Suez Canal get blocked up by a big barge, I mean, those are things that you can't really bake into a model per se, right, that's a black swan event that hopefully will never occur again, probably it will. But yeah, it's really just a whether some of those shocks.

And so again, it's case by case, it's nuanced, and what I always encourage our brand managers and we also talked to a lot of sellers in this space, who are using utilizing our tools or who we have, you know, close connections with. I always advise, at the end of the day, you the business owner, understand the nuances in the context of your business better than anybody else. And so while there are great demand planning solutions out there, of course we've released one as well. At the end of the day, any software, any model, it is not permission to turn off your brain, it is not permission to ignore the context that you know to be true.

And so, again, if you have working capital constraints, if you have physical limitations, if you've got Amazon imposed caps, if you've got category headwinds that are coming your way, you need to layer those factors on top of that sort of baseline level of forecasts in order to make the, I won't say the right decisions, because I don't actually think there is a right decision, they're just shades of wrong, and you're trying to be as minimally wrong as possible at any given point in time.

Ted Boeglin: That is actually a beautiful segue. So, Pratap, we've talked a lot about how you help companies make decisions through leveraging data to create better predictions. And one of the things that we've talked about is how, you know, you have to create probabilistic models that create different scenarios. And the importance of keeping people in the loop on the decision making. I know that one of the tools that you use is simulation, and that you create sort of simulated worlds. Tell us why you use simulations and how you help intelligent people make better decisions using your forecasting tools.

Pratap Ranade: Yeah, absolutely. I think for us, a huge reason behind simulation was rooted in sort of a lot of what we've been seeing in the last year with kind of unprecedented volatility, right? We went from like relatively stable times to extremely uncertain times and you had a lot of us just challenging the question of like, can you just take a model that fundamentally is trained on something that happened in the past? And use that to forecast the future? And obviously, because we've all seen, like, there's limits to that.

But does that mean that you can't really use data and machine learning. That actually doesn't mean that's the necessary corollary from that. And so, we asked ourselves and a lot of the team, myself included, we started our lives as physicists. So we're used to thinking about things bottom up instead of top down. And we said, well, at the end of the day what's happening, right? I as a consumer, I'm actually, yes I'm aware, I read the news about CPI, but that's not actually affecting my decision making. I'm walking into a store, I'm looking at products, I'm seeing how they're priced, and I'm making a decision based on the context at that time for me at that day.

And a lot of those interactions actually are knowable from past data. There's almost like these forces of like, cross product elasticity where certain products substitutes, you know, certain people they live in certain areas and so you have different kinds of types of people living in certain places, they're not suddenly moving across the country. And so there's a lot that you can know about these little knock on effects or butterfly effects. And so for us constructing that sort of like, atomic picture of the world that like the customer product and store level, how these pieces interact.

Actually created the foundation for this simulation. And where we sort of really got excited is working with our partners who where, again, on the demand and the supply side at companies, there's a lot of rich expertise and things that they've seen before. But if you can now marry the human and the machine, you could say, hey, I think I'm going to have stock out issues on the West Coast over the next X months, probably going to be between 60 and 80% of where you expect it to be.

Normally, that might just be an excel top down model. But actually if you can feed that in, that's not a uniform 60% everywhere. And some of that is normal, where are you still going to have access, where are you still going to have shortfalls? And so we found that to be really useful. So it's almost like a, think about it as a level builder for a video game, but you get to do, you get to be the game designer. And so that's how we've been finding stimulation to find utility and the supply side of the world.

Ted Boeglin: Pratap, what are some of the most common variables that you see, you know, very sophisticated companies who are doing as well as they possibly can to forecast demand and plan inventory. What are some of those big variables that they're looking at that have help them improve the accuracy of their forecasts?

Pratap Ranade: Yeah, it's a good question. I think a lot of the shocking secret is, it's not necessarily a huge vast array of them, it's like the quality of them and the granularity of them. So what we focus on is getting kind of the basics right, at like a granular and high frequency level. So can you get daily SKU level, store level price correctly? Can you get that a little bit up the chain, so like the price to the retailer, or the price to the consumer. Can you get some of the other factors that do drive behavior, which is particularly on like, if products are advertised, so that marketing spend.

So a lot of the controllable factors are things that are knowable, like volume sold, price, location. And then there's some simple things that we add for contextual data like demographics, a lot of the industries we work in, whether actually does matter a lot local events and holidays, matter a lot. So we do pipe that in as well. What's been interesting is actually given, again, if we you take a side tangent, some of you who are AI nerds have probably seen a lot of the development for language models and image models, what we found is actually weirdly adding in unstructured data from an image, stranger can help you spot like, substitute products. And we know this all the time as consumers you walk down the aisle, you look at deodorants, you can look at the design and you can kind of get a sense of which one's marketed to you.

So recently, that's been one of the things that we sort of brought in. And specifically that's been helpful with products where they're newer products, there's less sales data, just to try and make a prediction off of very little data. But I'd say, the core attributes are getting the sort of basics right, updating regularly at a very like atomic level.

Ted Boeglin: Yeah, it's interesting for you to say that it's not like there's some set of keystone variables that we probably don't know about, that everybody should be incorporating, it's more about the frequency quality of the most essential information that helps to improve that model. Alex, I'm interested to hear your perspective on that question to, what are you finding to be the most important pieces of data that you're feeding into your model that helps you get sort of high fidelity decision making?

Alex Kopco: Yeah, it's, this is not a cop out answer. It is exactly what Pratap said. One of the layers that I might add to this is, I think most folks in this space are aware of their annual seasonality. I peak in Q4, I peak in the spring, or whatever it is. But one of the other seasonal components that we think about is market based cyclicality in whichever locale you're operating in. And so a great example is we own a an at home fitness brand. And as Elon Musk among others predict that we may be entering a recession soon, one of the really powerful levers, layers, factors variables that you can add to it is this notion of a recession, and then you can try to get smart and leverage AI how deep will it be, if it's this deep. And this comes back to the question of simulation, what Pratap and his team are really focused on too is like, how do you try to predict them? Like the impact that a recession, a mild recession, a moderate recession, 2008, like crisis level recession. What impact is that going to have on people's discretionary income. Will they cancel their gym memberships? What happened in LA, but how many? And to what extent? and those people who do cancel their gym memberships. Well there's another confounding factor here which is, we just came out of a global pandemic in which 100% of gyms on planet Earth shut down at the same time.

So the question is, are people who canceled their gym memberships in a recession, gonna go re-up their own fitness gear? Or are they just going to dust off what they bought in the pandemic? So can we expect a similar lift in our at home fitness brand for net new products? Or not? What is the resale market going to look like? So I think there are innumerable new factors and new contextual specific variables that we would want to be able to consider that the power of AI and machine learning is pattern recognition, pattern application, and the more granular you can be over a longer period of time, the better the machines can be at spotting those patterns and translating them into a human context that allows us to understand it more efficiently.

Ted Boeglin: So you're both doing this in a very sophisticated level. I am willing to bet that a lot of the audience is looking at this and going well. I can't use machine learning or AI, and I don't have a big team that's gonna model the recession impact on home gym equipment. So, another way that I like to think about it is, what are the phases that companies should go through as they're starting to mature to gradually get better at demand forecasting and inventory planning. So I know both of you have also existed environments where like, a spreadsheet is pretty sophisticated for where some companies are at. So help me walk through how you envision the phases of map of maturity that a company goes through and demand forecasting and inventory planning.

Pratap Ranade: Yeah, so I can jump in and start. I think what, you know, refers to Excel and Spreadsheets, they are one of the greatest software applications ever. So I think, incredibly powerful, incredibly flexible. And I think like Alex had said earlier, like, nothing is an excuse to turn off the brain. So spreadsheets is tools for the brain, I think for sure.

3 Phases of Company Maturity

What we simply characterize them into three phases that makes it easy, but the first one is almost the data foundation. So even just getting all of that information into a clean spreadsheet is kind of a beginning, and being able to say, hey, if I'm, you know, you look at like Google Sheets is actually scripting and formulas and API calls that you can make from Google Sheets. In fact, one of our simulators you can actually call from a cell via an API call. So like, you could use a simulator straight from Google Sheets. So it's like, it doesn't need to be a massive IT undertaking. It doesn't need to be overwhelming or intimidating. I think that's the first thing is there's a lot you can use from simple tools today.

And I think that's sort of the data foundation is just getting it all kind of labeled getting a catalog. But that might start as a set of network spreadsheets that might move into databases, or data lakes. But just even simple things like making sure that the pricing and revenue management team, the marketing team and the supply team have access to what each other is doing. It's surprising, but you might have someone undergoing a big promotion in an area where there's going to be a supply shortage and we just wasted more money. So it's sort of the data foundations.

3 Phases of Company Maturity

The second step is really decision assistance, which is moving to great, can I get suggestions, can I get some alerts, can I get some sort of intelligence here. So I can basically as a human cover a wider scope, but be directed to focus on where it matters most. So some sort of decision this is tooling, and of course the promised land, which is our whole goal as a company is getting larger and larger tracts of these decisions to become fully autonomous.

3 Phases of Company Maturity

And again, fully autonomous, it's not everything runs by itself, it's actually surprising how much thought would need to go into, oh, the strategy I want to run autonomously. And so actually, we find a lot of that nuance becomes really interesting and really creative. And the person is actually setting up detailed guides for what they want the machine to do. And then it shifts to passive tracking and where I want to override and improve, but it's like, we think about it as these sort of three steps on the journey.

Ted Boeglin: That is a really helpful mental model, sort of start with the data, then move to decision assistance and then get closer to decision autonomy, even if you can never get to fully autonomous decision making.

Alex how, as someone who has done this process for many brands over time, how does that mental model resonate with you, and what would you add on to protect characterization of the different phases?

Alex Kopco: Yeah, I think Pratap did a great job summarizing sort of those three spaces, I'm going to use one of the analogies that he used to answer the previous question, which is, this notion of it's like being a game designer. And what I'll say is, most of the tech that we build, we prototype using a spreadsheet. We start with an incredibly simplistic use case. The very very, again, the quote Pratap, the atom was a fan of the use case, what is one SKU with the longest period of data that we want to get right. And literally what we have done is we have a data science team, and they come in, and they try 50 different ways of forecasting that single case in, that single product going forward. And what they will do then is land on, hey, this is the model, the algorithm, the variable that in the most simple use case yielded the best result.

And from there we then asked, we later on, we so you can sort of think of that as like, we've like, built the, we've drawn a circle in the game. And now we're going to start to figure out what is the terrain going to look like, is it a temperate? Is it not temperate in other mountains? Is there an ocean over here? And so you start to layer on these additional factors, and so then we say, okay, this model does it hold true for multiple products with different seasonal curves. What about multiple brands? What about multiple product categories? What about different channels? What about if this product only has six months of data? What if we haven't launched the product yet? And so suddenly, we start stress testing this notion, again, in Excel, on different use cases before we ever actually write a line of code.

And what we, why we do this is because it helps to prototype this stuff out. It helps to sort of wireframe it. So we're not wasting a bunch of mental capacity time trying to code these contexts like no program. And I think this, there's like this belief that you know, you, Apple is so good at making products because they're so intuitive, you can almost forget that there are tens of thousands of people spending millions of hours conceiving of that experience. And that it just, each one builds upon itself. And so it can be like AI and ML, they can sound like scary buzzy things that are totally unapproachable. But they're not, they really are built upon the foundation of Excel of data modeling. And something that Pratap said that struck me is this notion of the decisioning. I mean, that's conditional formatting, right? Like the most basic use case of that is, hey, I've got clean data in Excel, I slapped it into a pivot table, and I want to see what my seasonal trend looks like. And you do that little conditional formatting and it shows you the red stuff and the green stuff, that's just helping tell you where your eyes should look, it's helping inform you of, her at this time it's low, that could be good or bad at this time, it's high, that could be good or bad. But it boils down to your context again and how you want to think about and process that data.

Excel is an amazing starting point or Google Sheets, pick your poison. It's an amazing starting point for helping you to make those decisions and get comfortable. And in fact, I'll just add one more plug. When we're developing technology, oftentimes what we'll do is we will release that prototype to our users, which in this case are our employees. And what we're part of what we're doing is, I can sit here with my data science team and we can come up with the models and think through all these different business application use cases. But until we put it in the hands of our users, even in an imperfect way, i.e excel imperfect, quote-unquote. We are really interested in what decisions they're going to make based on the outcomes of that, because if we again, invest all this time and money and energy, now codifying these super cool, awesome models into a piece of tech that no one uses, we're no better off than if we had nothing at all. And so it's it, I would strongly encourage getting comfortable with thinking about this in those small bite size, sort of packetable starting points in Excel, and whatever you you think about and understand the use case for what you actually need that data, because it's just going to make it that much easier to drive adoption, that much easier to bake it into your business process and make decisions with confidence based on that.

Ted Boeglin: Yeah, I think it's been very helpful to talk through some of these examples, because I actually think it is very approachable for all companies to start making improvements in this way, even if it's Google Sheets or Excel and working up from there. But one of the things we've touched on as well is that these systems only work if they're sort of wrapped in a decision making process within a company that involves multiple people, multiple departments so that everybody is working in harmony together.

So Alex, maybe you can start, and then Pratap I want to hear how this plays out, also with some Arena implementations, but what have you seen is some really good decision making processes that are supported by these sort of core demand forecasting and inventory planning systems?

Alex Kopco: Yeah, we use our demand forecast, well beyond just inventory planning and sort of how much inventory to hold, we go all the way up the value chain. So we're thinking about we own a couple of brands that are heavily invested in silicone based products. And in October of last year, silicone, the price of silicone triple.

And fortunately for us, we had bought a bunch of raw materials prior to that. It was purely luck. We did not forecast for that. But we bought a bunch of raw materials ahead of time because we were able to take advantage of an economies of scale approach, because we knew that we were going to use that raw material, not just for a singular product line, but for multiple product lines across multiple brands. That helped us whether the cost heights. It also helped us as we sold through much of that inventory, to then go back to our vendors, for example, farther down the value chain and say, okay, we know that for the next six months supply we're going to need this amount, we're going to manufacture this number of widgets. So we could then negotiate on a bulk purchase as well to save on raw materials as they were purchasing it.

The other thing that we use it for, we use it in pricing, you know price is a real important variable that goes into a demand forecast, and like Pratap said, same particularly, you know cross price elasticity, understanding what the substitutes are with the complements might be. And so we use our demand forecast to understand the impact that our pricing, like how price sensitive are our products going forward, how price sensitive were they in the past, there's a lot to be learned, we encourage small tweaks in our pricing, particularly the beauty of selling on a marketplace, like Amazon, for example is you can make very very small tweaks, and they can have really big impacts in customer behavior.

That in turn helps train our model, and in turn helps us make better decisions so that we hold the right amount of inventory going forward. And we can do things like, reach economically optimal prices from a profit standpoint, right. And we don't have to say, well, our competitors charging this and this other competitors charging that, and our cost are this, and you're sort of doing this like bottoms up price, we can just come out with confidence, saying, you know what, we know day in, day out if we sit roughly at this price, here's our expected demand, here's our expected growth trajectory, and here's how we should think about our promotions process, our advertising process. And by the way, advertising is another way that we use it, if we want to really think about the split between our paid sales and our organic sales. If we're launching new advertising campaigns, we want to make sure that we're stocking up for that.

So again, it's we use this sort of systematic forecast as a starting point, for them layering in these additional sort of exogenous strategic decision based business, points of execution for lack of a better term. That again, the model might not be considering in its base form. But as we layer on top, we get a more and more accurate picture of what true customer demand is going to look like, probably. And that allows us to just make smarter buying decisions up and down the value chain.

Ted Boeglin: It sounds like it's kind of push and pull, right? It's like you develop this model, but then also the model helps then inform other decisions.

Alex Kopco: Right.

Ted Boeglin: Which then make sure you look back on what the forecast will be. So it does seem like it's a very iterative process that you're going through, which sort of starts at a rough estimate, and then refined, and refined, and refined as you cascade it through the other dependent decisions.

Alex Kopco: Yeah.

Ted Boeglin: For that particular product across the company.

Alex Kopco: It is, and the real meta thing here is that, as soon as you've used the model to make a decision, the act of executing upon that decision totally changes the model. And so then you need to rerun it, right? And that's where again, machine learning the power of machine learning the power of artificial intelligence, is that whether it be structured learning or unstructured learning, taking the action, better informs the model, and so you get better and better results over time. And so one of the things thatI coach my team very heavily on is, no action that we take no decision that we make is stasis. You don't do it, and then walk away and focus on something else, you do it, and you see the impact, and you measure the impact that, that had, because the world has just changed, you've changed the world in a very small way.

And so now you have to go back and see how that affects the future, how that affected the past. And so that can be, sometimes hard to get your head around until you see it in practice. And if you, again, like you can do this in Excel, just paste your next week's numbers in there and see the impact you change the world through that decision. So it's kind of fun, I think it's really cool. I like the video game analogy, because it's like you're on the next level now. And the level is totally different, but you've gotten smarter at playing the video game. And so you're better prepared to address the next level.

Ted Boeglin: Yeah, and Pratap, you're helping some of the largest companies anywhere, some of the most sophisticated companies anywhere go through these iterative loops. And I think Alex sort of described why it has to be iterative very well. It's, you know, your act of deciding actually changes, changes your future outcomes. How are some of the biggest companies using Arena and other tools to weave this into their decision making and get to that really fast iteration that everybody's looking for to keep their forecasts accurate?

Pratap Ranade: Yeah, that's great. And Alex great segue. I think the video game analogy continues to do actually work well, for so many reasons. And and I think one of the pieces that really resonated when Alex was saying is, if you do something you affect the world, you're like altering the world by taking an action, and so the world is not some perfectly measurable state that you could go and quantify and say, I now understand it. It's like you're gonna constantly like permute it. And I think what we see is there's sort of like two things we had to solve, that are really relevant, especially as companies get bigger.

So as companies get bigger, and most of our customers they're very large, right, very large international. And what winds up happening is a lot of them have grown through acquisition. And so you have actually what used to be many independent companies, of course, with the organizational silo and the data silo you'd expect. But actually from a consumers perspective, you may not know that the six brands you're choosing between are all owned by the same company. And so cannibalization is really important.

And so you sort of have to take on in a very first class way, this cross product interaction. And also, like you have to, there's also an amount of market conditioning you're doing each time you sort of take an action, like consistent stockouts or consistent discounting or price changing, you're actually reconditioning the market when you're operating at that scale, right. And so, all of these things interact, not just and then, the second part of it is, outside of your company. Now you have, you know, the talents here, we did a lot of like a work with the Defense Department. And the way they would call them is the thinking adversary, you have a thinking adversary, right.

So again, it's like, it's a competitor here. But the idea here is you have like, a competitor. So you have one layer, which is I pull a lever, my competitor reacts. Now, most of the world's like that scale is still run by oligopolies. And you've got like a few players moving determining the bulk forces in the market. And then you have minor adjustments that you can make around that. So for us, we found sort of a few things to be really critical here. So one is, as we're talking about simulation, and simulation, originally, a lot of the use cases for us were revenue management and pricing. But as we can tell through even this discussion that we're having here, that provides a data fabric that is useful for marketing, it is useful for supply. And like Alex was saying, what each of the big spend items that they be listed out are the same sort of spend items, which is, if I look at my holistic ROI, I'm expecting to get consumers to buy more at a margin. And the way I can eat into that margin or invest is, one is price. But one is the marketing levers, one is inventory? And so how do I actually use these? It's very much like problem, we formulate as a multi armed bandit, you’re just running around pulling a lot of these levers. And so that's been one component, which is the simulation.

The second part is, actually think about sort of the evolution of experimentation, right? So you think about one of the things that humans do remarkably well is, when I have no idea about something and I have no data about something, I'm going to make a reasonably good guess. Maybe not me specifically, but others who are good at supply chain would make a reasonably good guess. And that's something that machines don't do really well. But the question is, can you? So let's say you're coming into a market, you might look at your data, you might actually be a very big company with a lot of data. But it may turn out that every summer, you do a 50% off on Fourth of July. So no matter what model I put it into, I can't know your price elasticity, because there's one data point. It's like one sort of intersection point, right?

So how do you sort of intelligently start sampling, you can actually use the scale to start sampling. And again, we go back to the video game analogy, like any of you've played those strategy games like Starcraft or something, the maps, mostly black, and you go send a unit out there, and it explores learns a little bit about it. And so how do you choose when to try something new, and when to optimize? And I think the big risk with a lot of machine learning tooling today is optimize, optimize, optimize, but actually you're shrinking how much you know about the world. So you know that balance we found to be really important. So for us AI simulation, but two is it's almost this live real world experimentation. And what's nice about that is, as you learn, that basically updates your simulation. So you've got this sort of, like, you know, we're experienced was accumulated in a person, maybe the person leaves, you're accumulating that experience in a sin. And so that organization actually now has a copy of their best understanding of the world that's available when the pricing expert leaves, or the supply expert leaves, and that, we believe that's starting to create the initial signs of like a long term lasting, lasting advantage.

Ted Boeglin: Yeah, it's, you both talked about this by the way, which is that your people have to codify their understanding in some way. And I think that a lot of companies, especially small companies have this which is key person risk, right? You have one person at the company who understands it really well. And that creates a lot of key person risk if that person leaves or retires. And so you've both touched on this, but having that individual codify what they understand of the world into a system that can then go validate or disprove or ideally improve upon that understanding the world is such an important part of this equation, and something that you can start at any time. You don't need a database, a spreadsheet works just fine, for someone to record their understanding of the world, the key relationships, so on and so forth.

Okay, so we've covered some really big topics, we’ve covered supply, we've covered demand. But for everybody who has been through an economics class, they know that there's another output of a supply and demand curve which is generally price. And we have not talked at all about the current environment, especially with inflation, when you look at all of the indexes, the producer price index, consumer price index, consumption indexes, they're all going up in a very unattractive way, which is forcing brands to consider, do I need to pass increased prices on to my customers? And what does that do to my understanding of my demand forecast. And Pratap, I actually wanna start with you, because this is one of the places that Arena started, which was helping with pricing. How should companies be thinking about that really pivotal moment, which is, do I increase my prices? And run experiments or increase them durably, like how should I as a business owner think about that crucial decision?

Pratap Ranade: Yeah, it's a great question. We actually have an index of all the products that we track daily on their prices, and even you take like chicken breasts, and you look at chicken breasts in New York versus California versus the US average. And then you look at in different specific locations in Manhattan and Brooklyn, like this thing is actually, there's a widespread and what's crazy is you might expect that the rate is different. It's not just the rate, it’s the shape, there's actually periods of time where like, it's going down, it's going up. And so the interactions that micro level, when we roll them up, there's an interesting story. But what’s we see is really interesting is the way that companies and brands influenced the world is not my issue, I mean, they're not the World Bank, they don't issue a statement, as like, you know, as everyone on this call knows this, you're acting at the micro level, like great, I'm changing my price on Instacart, I am changing my price. Yeah, Alex you know much more about the Amazon world or in the store, right?

And how am I actually enacting that, and prices are not the same, right? Like when we work with consumer goods brands, the price elasticity to a change representative as free shipping, if you have over $10 off, it might be the same spend. But it's a very different reaction to like a price drop, or price to restore of credit, display matters a lot. So it's sort of that nuanced question of like, we fold it up into this elasticity word but like, when we do the modeling you'll have like display elasticity price. So these sort of individual components that sort of wind up affecting it. And I'd say my number one sort of, you know, or offer like a piece of advice for companies thinking about this right now it's like, now's the time to really try and understand the true nature of demand.

What is surprising is even with some of the most sophisticated companies in the world, there may be a prevailing belief, for example, that this product interacts with this product. And the belief might be so strong, but when you look at it in the data, and in price, which is usually the cleanest signal, the two products do interact in price, they follow each other closely and track. But when you actually, Alex mentioned a lot of the seasonal effects, when you count for these, and you subtract them out, and you look at sort of like, the demand residual that's affected by your price change or your display change. That interaction is changing. And again, at this point I can only speculate by imagine we live in a world where like, I'm being targeted via Instagram and YouTube and Tiktok. My preferences are changing faster than they were before, layered onto an already volatile macro environment.

So prevailing beliefs, even as recently as a year ago about which products were substitutes. They may not be true and sort of rerunning, or rerunning the analysis with the current data, and digging it a granular level, that's sort of like, definitely step one, to really formulate this changing view of elasticity, it's no longer enough to sort of do a conjoint and a segmentation study every six months, right? It's something that kind of needs to be done at a granular level regularly. And so I think like that deep understanding, the lucky thing is, it is in the data, it is in the data that companies have, and a lot of the signal is there. But I'd say that's the main thing is sort of like resetting, and challenging being willing to kill some of the sacred cows around what we might believe to be true. And the data might tell a different story.

Ted Boeglin: Yeah, so it's, I mean, summaries it's too complicated to give a single answer for how quickly the world changes and all the different interaction effects. But what it sounds like you're actually recommending is, is looking at your data, testing your prior assumptions and actually experimenting with different price points or different types of promotion to see what the actual effect is to expand your understanding of your true demand for that product. So that's, I appreciate you Pratap, because you give such thoughtful examples about what is possible.

Alex, I also want to talk about what is practical, because a lot of the companies we talked to were like, yeah, but I can't just experiment with my price every week, I've got a set of stable customers that are gonna go crazy if I'm constantly tweaking my price points. And so maybe you can take the practical lens on this too, which is, how do you consider whether or not to pass on increasing prices, or how do you view experimentation and getting that balance right as someone who's managing thousands of SKUs, and forced to make these decisions daily?

Alex Kopco: There's nothing to it, but to do it. It's one of those things where I've really grappled with this because in a world that I and my company live in where we operate direct to consumer sites, we're on a number of retail.com, we're on Amazon, we're a third party merchant. We have a lot of contexts in which customers shop, they interact with our products, we have brands, and so we want to maintain some consistency across all of those relative brand interactions. But I completely agree with Pratap which is, is it we are finding it time to test our assumptions. And I think, particularly for companies that have been around for a while, brands that have been around for a while, in particularly in a macro environment, let's say we enter into recession we're in this like inflationary period. Most people that are in their current job weren't in that job. The last time we had a recession, the last time there was a severe inflationary period.

So it doesn't hurt to put on your history hat for a minute and just look back in time at your industry. Or if you have it at your company, what happened then? And what are some of the market conditions that are similar that may at least help me build a mental model for what could happen now as a starting point. And through that lens, are there commonly held assumptions that you can challenge in a small way to get a signal for whether or not that change is actually something that could be sustainable? I think some of the best companies that come out of recessions, that come out of macro economic downturns stronger are those that do a couple of things, number one, they listen to their customers, and they have a great relationship with our customers. Like if you can't run a price experiment and you mentioned I have this stable, like cohort of customers that are super loyal to me, can ask them? Can you ask them how much they'd be willing to absorb? Right?

I mean, if you have that level of a two way conversation doesn't hurt to ask, because everybody is aware that times are tough, times are getting tougher. So that's one option. The other option is, are there segments of customers that you can test with? Can you target customers via Instagram exclusive, if it's a consumer good? Can you target customers, your loyalty members? Hey, loyalty members, we want to offer you the opportunity to lock in your current price now for a duration of time, because prices are going to go up for a non loyal segment, something like that, right? There are options, many options to test different segments, test different pricing models, pricing strategies, and just those, I think the no sacred cows mindset is right. It's an opportunity to really start to rethink, reshape yourself. And the outcome may be that you end up building an even better business than you had before the economic downturn, before the challenges that you face. Because guess what, as we said before, the world is changing every single day and taking an action changes the world. And so if you can change the world for the better in the context of your business, in the context of your customers, in the context of your strategy, that can make you better off and we use times of great uncertainty to as an opportunity to test things that we are ourselves uncertain about. Rather than, I mean, I think this, when the going gets rough humans we sort of trapped, we sort of have a tendency to regress to the mean, we retreat to the safe things. Can you push yourself and challenge yourself not to regress to the mean, but instead, again, I'm not saying fully overhaul your entire company, fully over hell your fundamental business model. But can you take small steps, try small experiments to get yourself a signal that might give you a clearer view of tomorrow than you have today.

Ted Boeglin: I think it's a really beautiful note for us to end on. I mean, you know, I'm a firm believer that the winners are forged in the hard times. It's hard to differentiate when things are easy. And Alex, I think what you were speaking to there, Pratap what you were speaking to was, the advantages of being in hard, volatile uncertain times for those people who lean into it and are willing to turn that uncertainty into an opportunity. So I think that is a really excellent sort of closing note for us. I want to give a very sincere thank you to Alex and Pratap for sharing their infinite wisdom with us on this topic, if like we could have talked for another hour on this because it's such an important topic, but that does conclude our webinar today.

I want to thank all of the attendees for the great questions that were asked. If we were unable to answer them live, then we will follow up with you after the webinar. We will drop a copy of the slide deck and an access to the recording via email to you tomorrow as well. If you enjoyed this content, we also have a newsletter that's focused on the most important topics for E commerce companies, we call it Smart Commerce, you can subscribe to that newsletter using the link in the chat.

Ukraine Crisis: How You Can Help

And before we say goodbye, I would be remiss if I didn't mention that flexport.org is raising money through donations to send shipments of relief supplies to refugees in Ukraine. It is some of the most important work that we do at Flexport. We will drop a link to, in the chat as well if you would like to learn more about these efforts or if you would like to make a donation. And with that, I want to thank all of our attendees for joining us today. Goodbye everyone.

Share the Video


Ready to Get Started?

Sign up for a Flexport account or ask to see our platform in action.