Deep Dive: Why You Should be Paying Closer Attention to your Sizing Approach (Hint: it's Not GLP-1s)

0:00The core drivers are the classic kind of inventory drivers, right? You get the wrong product in the wrong place at the wrong time, and then you either run into stockouts or you run into overstocks. I think the number is something, I think it's 1.7 trillion is the latest cost estimate in retail globally on the cost of stockouts and overstocks.

0:25Right now, everyone wants to blame the massive uptake of GLP-1 weight loss drugs for messing with apparel sizing and demand. That's an interesting theory, but it's mostly a distraction. The real story is that size planning has been stuck at the bottom of every retailer's to-do list for years, quietly bleeding margin. Core sizes sell out, fringe sizes rot on the clearance rack, and nobody treats it like the money problem it actually is. That's according to today's guest, Anaplan's James Toikoff, who has spent much of his career examining exactly this problem. He co-founded Syrup Tech to improve demand forecasting and inventory management.

1:05And after Anaplan acquired Syrup in 2025, he is now the VP of product at Anaplan. Today, we're getting into the real cost of getting sizing wrong, what has changed on the data side, and why Smarter Sizing might be retail's most underrated sustainability play. Let's get into it. James, thanks so much for joining us on Retail Remix. Thanks for having me. We're here today to talk about sizing and demand planning. Retailers pour so much money into planning, assortment, allocation, and yet sometimes it seems that sizing can not be the biggest priority. Why do you think that sizing is sometimes overlooked?

1:53Yeah, it's typically not part of, you know, the core stack if you think about, you know, assortment planning, merch financial planning, allocation is like the core, you know, three-legged stool. A few reasons for it. You know, one, the complexity of it. Sizing is super, super granular. So, you know, if you go down to a size store level, it's incredibly difficult and incredibly vast data volume. So, you know, it can be pushed to the it's too hard pile. Historically, we've seen huge data challenges here. It's really difficult to get clean granular data in order to make accurate size level predictions and decisions.

2:29And then I think there's just this overall, you know, We call it the good enough quote unquote fallacy, right? That you as a retailer, you just rely on historical size curves, right? And that feels like it's good enough, right? The reality is they're often inaccurate and they lead to margin erosion, but you feel like it's good enough. And then maybe the last one that I mentioned is just a lack of ownership, right? Like where does sizing live within the brands that retailers we work with? Sometimes it falls within merchandising, sometimes in planning, sometimes in allocation, sometimes in its own team.

3:01So this was this lack of clear ownership and accountability and put all of those together. And that's why, you know, sizing is overlooked sometimes. Has sizing become harder or easier to plan for in the last decade or so? I think there's arguments in both ways, to be honest. I would say on balance, it probably has become more difficult. And, you know, the reasons of why it's become more difficult is, you know, skew proliferation and, you know, the complexity of sizes. sizes have gotten all these different like medium long or different pant sizes in like three or four dimensions extended sizes across different size runs so there's there's more of it in more shapes than we had previously which makes it harder and you couple that with overall compressed trend cycles and volatility in the fit preferences right like say 10 years ago or 15 years ago So we'd already seen the beginning of this trend shifting, but still you could overall say there's still these three to five year macro cycles.

4:03I think particularly with the rise of social media and particularly with short form social media, the TikToks and Instagram Reels, the trends are just compressing, which means that coupled out with the rise of fast fashion, ultra fast fashion, that these trends are moving quicker. And that means that what used to be fine to think of as like, you know, size curves that are fairly constant, that's not the case necessarily anymore, right? And also be remiss, obviously, to say that we're seeing accelerated demographic and body proportion shifts, right? The whole story of GLP-1s, and we'll get into some of that.

4:37I think some of that is more fad that it's actually real sizing trend. But it is, of course, true that we are seeing in today's day and age body shapes and body proportions changing more rapidly than they potentially were previously. And I think all of these things mean that it's harder. It's also worth saying, however, that it has become or is becoming simpler in some ways, right? And the availability of the granular multisource data is becoming a lot easier. You know, 15 years ago, it was really hard to get the data. today with the vast majority of modern brands and retailers we work with, they have the data infrastructure in place to be able to get better at sizing.

5:19And then there's also just a broader intelligence evolution, right? The granularity is really hard when you have basic statistical models, when you're able to leverage sophisticated neural networks and advanced methods, you can actually make sense of it, right? So even if the problem's gotten harder, we have better data, we have better tools to make sense to that problem, if that makes sense. That does make sense. I'm so glad you brought up GLP-1s because it just seems like it would be impossible to talk about sizing without bringing up this really rapid uptake of weight loss drugs. And I see so much in retail reporting about shifting body sizes, people return rates increasing, and just, you know, it has really altered the retail landscape, including planning.

6:09Tell me more about that. Do you think that GLP-1s are having that great of an impact on sizing and planning? I think this is a really, really good question to be asking, right? How big of an impact is this really having? It's a great story, right? Oh, yeah, size curves are shifting everywhere. And, you know, it's been really hard for me. And the reality is in the data and the retails that we know, that we work with, it's hard to be able to, in an intellectually honest way, say that we're seeing permanent population-wide shifts, right? And in fact, you can see that even within like GLP-1 research, right?

6:47Like only about a third of GLP-1 weight loss participants, you know, remain on therapy after a year, right? It's concentrated to a certain particular demographic, right? in the US, I think it's something like 11 to 12 % of adults, which are heavily concentrated in specific income and age brackets, right? In general, in general, slightly older, slightly wealthier people are doing this. So it's hard to say this is really a population impact today, which means that you then, you know, you should really make sure that you're not just taking a headline news article and adapting your size. In fact, it can be a distraction, right?

7:30from the actual sizing problem. What could you be missing if you're focusing too much on the GLP-1 headline? Yeah, and with some of our customers, we're seeing, in fact, that larger sizes are doing better than smaller sizes, which you would think, oh, that's really counterintuitive, right? Or you think that potentially you need to cut larger sizes from an assortment, which might not actually be the case because the population that you're targeting is seeing slightly different trends, right? So seeing or taking on the narrative that GLP-1s are the primary driver, I think that just overlooks, you know, consumer buying behavior that we know, right?

8:04For example, in online apparel, you know, consumers cannot try items on before buying them, right? And so, you know, free or frictionless return policies, they remove financial risk from returns. That's completely independent from any GLP-1 messaging. And so I think the core takeaway and the core, you know, the core action for brands and returns for leaders there is make sure that you're strategically align in your sizing as opposed to over-focusing on narrow trend chasing. It's similar in my mind as like over-rotating on a single social media trend, right? Yes, it might be that this carries forward, but there's all of these other trends that are also occurring, right?

8:47Which means that just taking the headline overlooks the core problem. And that doesn't mean that it's not true that for certain parts of your assortment, for certain demographics, there is a move to smaller sizes. And questionably, there is some of it, but it's not, you know, the same across the board. And I don't think that that's the pure headline that you should, that you should focus on, right? And then you might ask the question, so what should you focus on? I think it's really the fundamentals, right? And the core ground up sizing decisions that will help influence and, you know, put your merchandising strategy into life, So understanding root causes, knowing your customers, prioritizing brand level consistency across components, and then, you know, modernizing the technology used for sizing to drive clear financial returns much more than, you know, being caught up in these headlines.

9:39What is the cost of getting sizing wrong? Have you sort of gone through the numbers on this? Yeah, huge, right? The core drivers are the classic kind of inventory drivers, right? You get the wrong product in the wrong place at the wrong time, and then you either run into stockouts or you run into overstocks. I think the number is something, I think it's 1.7 trillion is the latest cost estimate in retail globally on the cost of stockouts and overstocks. Stockouts, of course, you know, you're a customer, you walk into a store, you look for something online, your size isn't in stock. You then decide not to buy that item, but go to a competitor.

10:18That's, you know, a lost sale that you're seeing. The inverse is, hey, I have way too many sizes. Say I have way too many XS sizes and not enough customers that are buying that. That means that we have capital tied up in that inventory. There's the actual financing costs of that. There's kind of the cost of that product eventually having to be either discounted or going to outlets. So you have all of that. And then on top of all of that, you have like the erosion of customer trust, right? Whether that's inconsistent or it's unavailable sizing, it creates friction. And we all know this, like, here we go.

10:52You know, we really, really like that product, but we can't find the size. That's really frustrating and creates like a long-term value problem. Or I typically buy size, you know, L, but, you know, suddenly L is really, really big for me. That's also really frustrating. So there's a whole bunch of costs associated with it. For sure. Yes, I am six foot one. And so I feel like a large portion of my shopping life has been about sizing woes and always feeling like I'm on the hunt for the brands that understand me and the sizes that fit me. So on the flip side, for those retailers that really get sizing right, I think I can imagine what you'll say here.

11:33But what do they gain when they do get it right? Totally. I mean, it's on the one hand, it's the reverse of all of these downsides. that you're maximizing full price sell through, you're reducing your markdown liabilities, you're lowering return costs, you're freeing up inventory and working capital and probably overall increasing brand loyalty, right? To your point of, hey, I'm someone who struggles with sizing. Well, if I finally find a brand or a retailer that you feel that gets me, just creates a lot more loyalty, right? There's probably a few other things additionally that I'd call out of what you get when you get this right.

12:10One is, I think, a whole piece around strategic freedom for assortment breadth and depth. So when a size distribution is predictable, the merchants in your team, they can confidently commit upfront budget to more style introductions, right? Or higher margin, capsule collections. So there's a whole piece of like you're giving your merchandising teams or your merchant teams much better abilities to test and create opportunities for real upside as opposed to just protecting against the downside. There's also a broader piece, particularly in the US, around enhancing price integrity and brand equity.

12:46And what I mean by that is retailers avoid conditioning their shoppers on waiting for clearance sales on off sizes. When that's not the case, you just don't have to create as many of these bargain bin effects for your shoppers. But you can rather protect the full price brand perception of your brand by getting sizing right. And then finally, it's just being able to hyper-localize your assortment, right? Particularly, certain sizes run smaller, like pink sizes, for example, often run smaller than a navy blue size. Why? Because some kids or younger adults buy those sizes and they're just by nature smaller.

13:25So being able to do that also creates an ability to make sure that you've got the right products in the right sizes, really for the right location, for the right buyers in those locations. You co-founded Syrup before Anaplan acquired it. So you've really watched the data science mature on sizing up close. What has changed that has impacted size optimization over the past five or 10 years? Yeah, absolutely. And, you know, it's, I think the equivalent has happened in, you know, whether it's predictive AI or whether it's sizing technology, that has happened at the macro AI scale, right? Like probably less flashy and less like, you know, all-consuming than some of the advances that we've seen with, you know, the large language models and then the rise of agentic workflows and now kind of fully autonomous agents.

14:20Something similar to that has actually happened beneath the hood of what we're able to do to help brands and retailers with their decisions. The big underlying trend is that, you know, The core data today basically all lives in cloud-native model architectures. That means that whether it's our data scientists or our customers' data scientists, or you as an analyst, you're able to just pull that data in a lot easier ways. And so we have access to orders of magnitude of more data than we had even just a few years ago. Then the really, really, really exciting piece is the whole modeling layer. Even when we started Syrup in 2020, the vast majority of the industry was really still using some sort of auto-regressive or historical sales curve fitting.

15:08Maybe there was some basic statistical model, but really it hadn't changed much in the preceding 30 or 40 years, particularly on sizing. It was really just, what did you sell last season or what's an average of an average that we apply that to today? we're able to be orders of magnitude smarter i would say today right by being able to take things like attribute embedding so you know think of this as you have all of your different fit attributes or style attributes that you have in your you know product master data that we can pull from product imagery and we can take all of that context that we know about an item to just create a lot smarter you know representation layer that we can then do deep learning on right so we You can understand, oh, wow, you know, this product looks like that and that product looks like this.

15:57The sales of this product were that, you know, they were like that. And then you can put that all into this. Hey, we're getting a little like too technical at this point in time. But you can put all that into like fairly sophisticated forecasting engines that use these multiple towers that take data at the same time. Right. So you can process all these disparate signals simultaneously. So localized weather patterns, regional demographic shifts, marketing campaign velocities, channel specific return rates, and we're dynamically weighting that, right? And the output is just a much, much, much smarter size curve, right?

16:32The same way that like you can go from like, you know, chat GPT-3 to like chat GPT-3 5.6. And, you know, it goes from middle school maths to being able to solve some of mathematics most complicated problems that we've faced with the history of humanity. something similar is happening with kind of the underlying technology that we have today. That sounds so satisfying. Planners can sometimes be nervous, I would think, about handing their decisions over to a model or letting a model lead that decision making. How do you coach people on how to toe the line between letting the data lead and keeping a human and control of the final size mix?

17:13Yeah, 100%. I think there's probably two core components, or maybe it's three. One, I would say, super important kind of as you're embedding more intelligent, you know, mathematical models or AI models into your decision-making, that in that implementation phase and in that, you know, rollout phase, planners are super, super involved, right? No one, I think, is going to trust a black box. Really, really hard. Because if you've done this job for many, many, many years and we've seen season in and season out, you want to understand what's going on. And so as you're bringing on new technology and new tools, you need to be involved in that implementation phase, which also means that you can be in the loop, right?

17:56That's both in the loop during implementation, but it also means as you're getting outputs from this engine, you have the ultimate decision-making authority. And, you know, I'm a huge believer that particularly in retail, this is super important. That, you know, advanced algorithms, they can provide a really strong, mathematically sound, optimized baseline at whichever level of granularity you want to. But at the end of the day, the planner continues to be the pilot of this, right? And that there is immense value in human expertise to steer both the strategic intent of the model, but then to also be able to set up certain guardrails and make certain changes because you just know better than the model.

18:39I think that's probably part one and two. Part three is I would say that this is more of a mindset piece for planners, but that they're able to shift from I need to be involved in every single detail to I can manage by exception, right? I think this manual, line-by-line, spreadsheet adjustments that we see many sizing teams and planning teams do, I really think that's something of the past. And, you know, if you really hold on to that, it's holding on to, you know, the fax machine or something, right? Versus being willing to, you know, take that baseline and, you know, doing spot checks, making sure that you're setting the strategic intent properly, and then focusing on the exceptions, focusing on the outliers, that any modern AI native system should be able to surface for you.

19:30So you're not looking for that needle in the haystack, but that system is presenting that needle to you on a silver platter and you're then expecting that needle. That's really where I think workflows are going and what I think planning teams should really internalize in order to make better decisions and to be able to combine the artificial intelligence and the human intelligence in the most effective way. I would think that smarter sizing planning would also align beautifully with sustainability. That's something we haven't talked about yet. Walk me through how getting sizing right could reduce waste.

20:06Yeah, 100%. And it's a super important topic, right? In apparel, for example, it's apparel and fashion. It's an industry that, you know, for every 10 items that are produced, only three or so are sold at full price and three to four of them are sold at a discount and then three end up being destroyed. And I think the latest industry estimate I saw was 2.5 to 5 billion with a B. Items of excess stock are generated annually. So it's a massive problem, right? And sizing makes this problem so much more difficult than it does for industries that don't have this size complexity. To answer your question, the three things that this does are where better sizing helps.

20:45One, it reduces overproduction. You're, by more accurately forecasting demand, and retailers and brands can produce few items that they're just unlikely to sell but that reduces the waste in the finished good but also in raw materials water and energy that are being produced to to create those um to create those items if you know that's number one number two you've minimized overall waste you create at the end of the season right so there's fewer unsold items that end up you know in landfills are being destroyed and then number three is you know it just overall lowers your carbon footprint, right?

21:19A more efficient, more responsive, a more intelligent supply chain with less waste means overall lower carbon footprint, right? Whether it's manufacturing, transportation, or disposal. And so there's, you know, the initial effect of, well, you just produce less things you don't need, but there's all kinds of these second or third tier effects of making this operation more sustainable. All right, let's end on a forward looking note. If a retail leader listening to this can prioritize one thing about their sizing approach in the next year or so, where do you think they should start? Yeah, maybe the answer is slightly different for whether if you have the infrastructure, if you don't, if you don't have the kind of data infrastructure that allows AI algorithms to kind of you know understand data and take some quality start there right like the data layer and the data infrastructure of the overall you know deterministic engine planning engine that you need to kind of as the core bases that should always be the starting point right so if you don't have that start there matters a lot if you do have that and i think evermore retailers are in a good spot i would focus on picking a category, picking a brand that you're managing, someone where you know that there are size issues, and just test this, right?

22:43There's easy ways to benchmark, you know, more modern tools versus what you're doing today to see like the incremental value that can be generated. And sizing is a really nice one because, you know, you don't need a broad overhaul of your entire planning suite, right? You don't need to rip out your existing, you know, PLM or your existing, like a sort of planning solution, but it's an easy one that you can like plug in and really start seeing value immediately in something that, as we discussed at the beginning, is heavily overlooked. So if you have the data infrastructure, I'd say this is a good moment to move away from old historically static size curves to something that's more modern, more tailored, and that will just produce better results.

23:22Yes, and also make more clothes for tall women. Exactly right. Well, thank you so much for joining us, James. That was super informative. Lots to chew on here. Thank you. Thanks for joining us for another episode of Retail Remix. Be sure to subscribe so you don't miss a future episode. You can find us on Spotify, Apple Podcasts, YouTube, or your favorite podcast app. Until next time, keep mixing it up.