The Demand Edge: Smarter Forecasting for 2025 and Beyond

Paul Sonneveld
I'm your host, Paul Sonneveld, and today we're talking about Amazon vendor forecasting and demand planning. If you ever wrestled with purchase order volatility, debated which data signal to trust, or struggled to align buy plans with production and cash flow, this hour is for you. 

Now, before we jump in, just a couple of quick updates. First, Amazon is coming up on the ninth and tenth of December. It's a two-day virtual event with practical insights to help vendors win in 2026. Early bird registration is open now and free for Amazon vendors and MerchantSpring clients. Just scan the QR code on your screen or head to merchantspring.io to grab your ticket. And of course, Amazon Unboxed is also right around the corner. Luke and I will be heading to Nashville in early November, and we'd love to catch up with as many of you as possible while we're there. If you are attending, send me a quick message on LinkedIn so we can connect in person. 

All right, let's get back to today's topic, forecasting and demand planning. Now, to help me unpack the topic, I am joined by Nathan Grimm, founder of Mercantile Commerce. Nathan spends his time in the trenches with brands, helping them work through forecasting frameworks, inventory risk, and the realities of Amazon ordering behaviour. He's here to share practical approaches you can put to work immediately. Hey, Nathan, it is an absolute pleasure to have you back again today. 

Nathan Grimm
Thanks for having me back on the show, Paul. I'm happy to be here. 

Paul Sonneveld
Now, just before we jump in, a quick reminder to our audience that this is a live show. If you're watching us on LinkedIn or YouTube, go right ahead, use that comment section, or, you know, I think it's both comments on YouTube and LinkedIn and pop your questions in there, and we will try and get to them during the course of the show. All right, let's get into it. Nathan lets me kick off with this question. Why is demand planning now mission-critical for Amazon vendors? In other words, why are we even talking about this? 

Nathan Grimm
Yeah, I think back to 12,13 years ago when I started in this space, and it wasn't as important because Amazon tended to be a much smaller percentage of a brand's overall sales. So as that share has grown, planning has become so much more important to make sure that inventory is available for the channel. And when one of your major channels is out of stock, you miss out on a bunch of sales. And if you overforecast and have too much inventory, then all your cash is just sitting there in the warehouse, and it makes you less nimble. For those reasons, forecasting on Amazon it's always been important, but it just keeps getting more important as the channel grows and as margins become more and more important for every brand. 

Paul Sonneveld
Now, when it comes to forecasting, there are a lot of ways of doing it and a lot of frameworks, a lot of approaches. What forecasting frameworks do you like or you know, like is probably the wrong word, like actually work in your experience and maybe because we're in Q4, specifically in sort of volatile categories or seasonal categories. What are the frameworks that you like to hang your head on? 

Nathan Grimm
Yeah, that is a good question. I think the first thing to think through is just what is actually happening when you're doing a forecast, and where all the different stages that a forecast goes through. When somebody says demand forecasting to me, I actually think about something very specific and probably much more specific than the average person means. A demand forecast is really just looking at the data of the sell-through data of a brand's products and figuring out, okay, if I was in stock all the time, and everything was normal, how much would I expect to sell? 

Now, that gets taken by a lot of different teams and gets turned into a lot of different things. The supply teams goes and makes manufacturing orders and shipment plans. The finance team goes and advises the board on how much they think they're gonna sell, what they think margins will be. And then the team sets targets for everyone, for the marketing team, for the sales team. And so I think having a clear view of all those parts of the process and not trying to mash too many of those things into one step is probably the most important thing when you're thinking about frameworks. 

Paul Sonneveld
So I just want to clarify a little bit more, Nathan. Are you talking about sort of how the different forecasts link together and how they flow from one another, rather than sort of building in one overall encompassing thing, having people build different signals to pass on through the chain of the organisation? Is that what you meant? Or did I miss something there? 

Nathan Grimm
Yeah, I think, yeah, thinking about how that forecast gets used by different parts of the organisation is probably the first thing I think about when I think about frameworks. 

Paul Sonneveld
Yeah.

Nathan Grimm
Because if the forecast says we're going to sell a hundred units and in December, and the supply team is like, great, I will make one hundred units. They'll be here in December. And then the finance team says, Great, we're going to sell one hundred units. Well, if something happens in that shipments delayed and I have zero units, then everything comes out the window. And obviously, in reality, it gets a lot more nuanced than that. But if everybody's working on the exact same number with no buffers, with no plans on like, okay, how are we actually going to get this done? And how do we manage our risk along the way? Then, reality ends up being much different than the plan in ways that can be predicted and managed for. 

Paul Sonneveld
Yeah. Yeah. Now that absolutely makes sense. Which sort of gets us onto some of the variables, right? In building a forecast, particularly how do you balance, obviously, a hundred units of a forecast is great in theory, but there's seasonality, there's lead times, demand shifts, maybe, purchase orders are coming in a little bit earlier because there is constraints around inbound logistics. So how can vendors sort of balance seasonality, lead times and demand shifts all at the same time? 

Nathan Grimm
Yeah. Yeah, that's where I think it's really important with the demand plan to focus in on a statistical model that can incorporate as many of these things as possible. And then use probabilistic forecasting. So what does that mean? It means that really you should be looking with any forecasting model worth its salt should be trying to identify things like seasonality. For instance, I work with a lot in the outdoor industry. And so things that you use while camping and hiking have predictably much stronger demand when it's warm, when the weather's nice. and much less demand when it's cold. Conversely, some products are for warming products, and so there's huge demand when it's cold and snowy out, and there's almost no demand when it's warm out. And so these things are fairly predictable, and any good forecasting algorithm and forecasting system will be trying to identify those predictable seasonal trends. 

But also, there's challenges with that because SKUs turn and demand changes. And so you want something that is using good statistical principles to incorporate recent, pay more attention to recent data than old data, while still keeping an eye on both so that it can track seasonality. As well as one that can look at things in a hierarchical sense and say, okay, well, this SKU is new, but it's very similar to these other SKUs, which have known seasonality and demand trends. And so what can I interpolate or infer from the category it's in from the parent product it's in, so that I'm not starting from scratch every time I have a new product. 

Paul Sonneveld
Yeah, so you're trying to pick sort of like SKUs that have exhibited behaviour that you'd expect from your new SKU and trying to model off that. Yeah, absolutely. What about, I just sort of want to talk a little bit more about, I guess, the uncertainty around kind of tariffs and lead times and things like that. Like maybe, you know, this year or maybe for next year, it's probably too late for Q4, but are you building in additional buffers or safety elements into your forecast versus what you've done previously? Are any sort of things where you've tweaked your approach and thinking on that? 

Nathan Grimm
When looking at next year, I think it's important to think about these things in terms of regression to the mean. So it's easy for us to look at the recent past and say, OK, the recent past has had a lot of uncertainty with tariffs. Look at COVID, where it had a huge demand shock or a huge supply constraint and demand shocks at the same time, that created all this uncertainty. But over time, those things tend to regress towards historical averages. So whenever I look at things that have happened in the past, I think, okay, are we through it? Is this likely to continue? Or is there likely to be some other thing that we don't know about that's going to be happening in the future? 

So there's been a big change that's been impacting supply chains or demand. What I look at in the future is not to say, oh, well, this is going to continue exactly as it has been. If I really am uncertain about the future, what I do is I take those things that have had a huge impact on the past and I ask, okay, what would things look like in the future if that impact regresses towards a more normal state of things? So when I look at the future, I don't really know what those things are going to be, but I do know that they won't be identical to the last year. And so really what I do when I try to understand the future is like, OK, how do I control for what's caused us a lot of churn in the recent past? And then what in over history through all these different periods of uncertainty have been healthy levels of padding to add to supply times percentages for buffer stock and things like that. 

Paul Sonneveld
Thank you. Yeah, if only we could predict the future with a 100 percent accuracy, then you and I would both not be sitting here, we'd probably be enjoying some sunshine somewhere else. Yeah. 

Nathan Grimm
Yeah. Yeah. I think it's tempting to say, okay, this, what just happened is going to happen again, or there's going to be some white swan event. If I can just predict it, we'll be okay. I think really a black spawn event. But really, the most probabilistic thing is that all that crazy stuff that just happened will get more normal, and some new unpredictable thing will happen. So really it's creating a plan that works in many environments as opposed to trying to figure out what the precise environment will be in the future. 

Paul Sonneveld
Yeah, and I guess part of that is just reviewing I mean, whether, whatever the right reference experience is, but like we're doing what happened last year, what are the things that would have put us in a better position to avert certain scenarios, and just trying to build on learnings, but in a more general way, as opposed to building for one specific, like what you're saying blanks on event. So, let's talk Amazon vendor. We log into the Amazon vendor console, right? There's so much data there around that you could use to build a demand forecast. You know, everything from retail analytics, shipped, ordered, revenue. We've got purchase orders. 

Nathan Grimm
Yeah. 

Paul Sonneveld
We've got Amazon's forecasting reports. Where do you start in building a demand forecast? And, you know, what are the vendor signals, Amazon vendor signals that you pay attention to? 

Nathan Grimm
Yeah, I always start with shipped units and depth of cover. It is tempting to start in a lot of other places, but really, what the commonality between every vendor and Amazon is they're trying to fill consumer demand. So the most, that root signal that's going to be driving both Amazon's behaviour as well as the vendor's behaviour is going to be how many units are consumers buying.

And then I look at their depth of cover because if Amazon has if I were Amazon and I had 26 weeks of cover, I wouldn't be buying inventory for a few weeks, I'd probably be waiting because I have too much, and if I had four weeks of cover, I'd probably be ordering more than normal. So it's important also to kind of keep your eye on how much coverage Amazon has at any given time, because if it gets lower than normal, it will probably be increasing. Their short-term orders will be increasing. And if it gets if the coverage is deeper than normal, their orders will probably be less frequent over the near future. 

Paul Sonneveld
Are you just on that? Are you aware of what Amazon's internal days cover targets are? Do you have a view on that based on at least some of the categories that you've worked on in terms of outdoor categories? Do you sort of have to put a stake in the ground there and say, I think Amazon is working to, I don't know, whatever it is, right? 60-day cover. Because that will help you determine when Amazon is likely to send out those purchase orders. Of course, there's lead time and all that. There's other factors that go into it. But have you sort of empirically, are you using sort of an assumption, either implicitly or explicitly, in your forecast? 

Nathan Grimm
I think there's normal ranges. So I think eight to twelve weeks is pretty normal. If I see them come in with an order that will put them up to, 20+ weeks, that's when you might want to intervene and say, even if I have these units, I might not fill these if I don't think that the demand is really justifying this depth of coverage. And that's really just to protect your net PPM with Amazon to prevent a possible markdown sequence in the future. 

And then if something is falling below eight weeks of cover, really, what that triggers is, okay, is Amazon profitable? Are they just not ordering very many things right now in general? But really, is there a problem with demand or with profitability that might cause Amazon to not like this particular ASIN as much as they did in the past? So, I hope that i hope that answers your question of you know normal eight to twelve, but in reality, it depends on the time of year depends on their cash position which we'd have no visibility to and it also depends on their strategic plans on if they think it's strategically beneficial to go deeper or lighter on coverage given the macro environment. 

Paul Sonneveld
Yeah, no, that makes sense obviously. There's that's you know, Amazon's yes, trying to fulfill demand but they're also trying to hit their working capital agenda for the year, whatever that might be, and physical constraints around their inbound network. So that works. And a view of future marketing events and activities that you may not have visibility to at that point, which also kind of gets me into, well, the part that you do have visibility of, right? So you do have visibility of, well, I guess not all pricing decisions, but, you know, certainly marketing support, whether it's promotional activity or deals you've agreed with your Amazon vendor manager or advertising campaigns that you're running on and off Amazon. These are all impacts the demand signal. Ultimately, how do you build those into your forecasting models, and how do you account for those? 

Nathan Grimm
Yeah, it's really important to pay attention to those things that will impact demand that Amazon doesn't know about. You mentioned a couple that are really big, which is deal planning and Adplanning. The other one, where Amazon just has no idea until they get the data, is on new products. And so the way that we account for those is different because those are all very different situations. 

For deals, we're just doing longer-range planning than Amazon. With Amazon, you put it in with enough time so that they can order up inventory, but you kind of wait until that last is as late as possible because things can happen with pricing to really scuttle your deal plans. But when we're looking at demand, we're putting out those deal plans or map break windows out for the full planning horizon that we have so that we're manufacturing and bringing in enough inventory to supply that. Now, with advertising plans, you always want to pay attention to are we going to advertise the same amount next year, a lot more or a lot less? Or have we identified a product or an important product that we're going to advertise a lot more or a lot less? Because that should impact demand. Same thing would go with price changes that are coming up. 

With new products, that's really where a lot of the art and experience starts to come into play. My expectations for a new product are super wide. You know, the number of results that won't surprise me is super wide, whereas for a product with multiple years, it's pretty tight. And so that's where you're looking at what's a comparable product, what category is it in? Is this going to be variated with an existing product or start out on its own page to try to look at what should we expect to start out from this product? 

Paul Sonneveld
Yeah, I think that is so hard on Amazon, right? Because I'm just thinking, like having spent many years in traditional retail, you know, new products you can put in a location in the store. You can almost sort of negate the other variables, right? Like, you know, I used to have another product in this location that behaved like this. I'm going to put the same product, similar things. I expect it to behave like this, right? But in Amazon, of course, a new product doesn't have the same number of reviews, doesn't have the same organic rank, doesn't have, like, there's all these other variables. So, you know, it must be very, very hard to kind of deal with that. It's probably a separate topic in terms of how you launch your product and how you give it your best chance. 

But where I wanted to sort of get to was around the promotional plan, right? So I imagine you've got your promotional plan, you've got promotions, deals all locked in, you've got a forecast, you've got a view of what's going to happen to your demand, right, based on that. What do you do if Amazon doesn't seem to be factoring that demand? They may not have visibility of the future promotions that you're planning to run, or they've not factored it in properly, and therefore they are under-ordering, or their demand signal just doesn't seem to live up to what you think is required. How do you deal, like how do you escalate that or how do you work through that with Amazon, really, to avoid the risk of really out of stock, right? 

Nathan Grimm
Yeah, with Amazon, I mean, the magnitude of the issue really will control how much time and effort you want to put in this. If this is a product where you think it'll sell a hundred units and they've got 80, perhaps the juice isn't worth the squeeze. But for big issues that are going to matter to the tunes of tens and hundreds of thousands of dollars of sales, you really want to use all the escalation points you have available to you. And you know, everybody knows that the bigger you are, the more influence you're going to have on the vendor side. So everybody can create a case can do some research into the SKU to see like, is there an underlying reason why Amazon's not buying enough? Are they unprofitable? Is there, has there been very little previous demand, are my expectations reasonable? 

And, you know, the bigger you get, the more you have, you might, smaller vendors might not even have a vendor manager, bigger ones could have a vendor manager, and even bigger ones still could have executive-level interfaces. And so really with vendors, the escalation path really varies with the size, the dollar amount you're talking with, because you're going to have a lot more access to those escalation paths, the bigger you are and the bigger that deal is. 

Paul Sonneveld
Yeah, makes sense. By the way, just for those tuning in, we've got quite a few people tuning in. This is a live episode. I can see that Jackson, great to see you, Jackson, has already popped a question in the chat. We'll get to that shortly. But yeah, feel free to keep your questions coming because in about two or three minutes, I'm going to put them in front of Nathan here. A couple more from me. Overstocks. right for whatever reason you may end up with way too much inventory in Amazon system. And obviously, POs are drying up, you may have more product coming in, how do you go about fixing overstocks? Probably, technically, outside forecasting but but it's kind of part and parcel of of managing the flow of inventory through the system. What are your tips here managing overstocks? 

Nathan Grimm
Yeah, a lot of that comes down to performance management. So step one is always looking for those on-off switches, which would be, am I losing the buy box? Which could be because of another seller or a competitor who's undercutting Amazon's price. Is Amazon unprofitable? Because they might just choose to start disengaging with products where they're not making money? Is the detail page messed up? Did my ads stop working? And so, really doing that performance management to figure out, OK, is it a traffic problem? Is it a conversion problem? Is it an availability problem? And how do I fix the underlying performance issue? 

Second, if there's no underlying issue that's standing in the way and is just a softer demand than planned, maybe macro conditions made less demands than everybody expected. Then it really comes down to price and price and traffic. You can buy more clicks and you can drop the price. And so with dropping price really, you know, deals of the day, best deals, lightning deals are your friends because those have a traffic driving component as well as a conversion-rate-boosting component of dropping the price. And then everybody can go buy ads. And so talking to your Ad team to see if it makes sense or if the juice is worth the squeeze elsewhere would be what you do to get rid of out of an overstock situation.

Paul Sonneveld
Yeah, I think really what you're saying is you've got to find a way to stimulate that demand signal. You've got to try and get that thing up so that the day's cover comes down, which then sort of hopefully kicks Amazon's algorithm into gear to start placing some further purchase orders down the road. Okay, last one from me for today before I open up to our audience. As you work with vendors and you have for a long time, what are the biggest blind spots and mistakes that you see Amazon vendors make as part of this forecasting process? 

Nathan Grimm
It really, they're as diverse as the company you work with, but some common themes, one would be focusing on sell-in versus sell-through. On Amazon, you have to focus on sell-through. Most every purchasing issue is going to be, well, there isn't demand for it or there's demand, but Amazon's not making money. So focusing on those consumer metrics as opposed to trying to convince somebody at Amazon to buy more is step one that gets you working in the right zone of where you can control the outcome or influence the outcome. 

And then second, is just education on the true drivers of that demand. There's a host of issues that could be preventing an item from selling, and then there's a whole other host of activities you can do to try to stimulate that demand. And so the more education on those topics, the better. Lastly, I think the other major blind spot is similar to the drivers of demand and getting more orders, is that one thing I see a lot is people will look at purchase orders. And they're like, Why isn't Amazon purchasing? Or using purchase orders as the basis for their forecasts. And you're just going to get a lot more noise because all of those Amazon-specific considerations, like they're open to buy, their inventory constraints are going to be clouding how much demand there really is and how much they're eventually going to bring in to satisfy that demand. 

So focus on sell-through. Build your demand forecasts off of those sell-through data points, first and foremost. And then just make sure like internally, you have really robust maturing systems around taking a demand forecast, turning it into a supply plan, financial targets, and then targets for the team really keeps you safe and helps helps everybody row in the same direction instead of the demand plan being a hot potato that nobody wants to take credit for because everyone will get mad at them later. 

Paul Sonneveld
Yeah, yeah, yeah. I remember my battles with the forecasting team and the trade planners back in my old retail days. And why do we end up with so much stock? Yeah, lots of fun. Okay, great. Let me open up to our audience. We've got a few questions here, particularly from Jackson. He's contributing a couple of great questions. Feel free to open any, any further questions if anyone else is tuning in here, but let's kick off. So Jackson asks, how do you perform demand planning? Any preferred models or services? So I think he's probably, you spoke about this sort of generically at the start of the podcast, but I think he's probably asking about any specific algorithms or I know there's softwares out there that really help you with this as well. What's your, how do you approach this question? 

Nathan Grimm
Yeah, in terms of pieces of software, I use a generic demand planning tool called SKU Science. It will forecast any demand on anything. It could be hospital visits. It could be retail sales. It could be anything. But so long as you have a time period and units that are happening, it will forecast that. It uses a different statistical model depending on what the demand is looking like. But it uses best practices in terms of when to use an algorithm that is better for low and unpredictable demand or intermittent demand versus high demand. It will weigh more recent actions above previous actions. It will look for seasonality. 

There's a lot of tools out there that do the same thing. In particular, I wasn't finding any focused on the Amazon, on the e-commerce space that really added value. And so I use that generic one. And then on top of that, I still do a lot of Excel work to really collect all the data I need, as well as factor in things like deals, things like, product transitions, things like changes to advertising plans. And so yeah, there's probably twelve or thirteen different sub-processes under demand planning in my world. And that's a piece of software that does a few of them. 

Paul Sonneveld
Thanks, Nathan. I've got one more here also from Jackson, with no vendor manager or born to run opportunity. And, you know, presumably there's no purchase orders there. How do you convince Amazon to create POs on new items? 

Nathan Grimm
You can drive traffic to the page if there's a buy box you can try to get people to buy it even though there's no promise date and then i suppose a hybrid model could be helpful for that as well where you stock it on the seller side um if even if uh vendor central is the platform you want to be selling that item on long term. 

Paul Sonneveld
You sort of yeah, you start offering the product, put an offer against an ASIN through a 3P account, generate some sales and reviews in the hope that Amazon's forecasting algorithms will kick in, see that's available in the vendor catalogue and start ordering it. 

Nathan Grimm
Yeah, if there's traffic to the page, they'll buy at least one or two. And then if those sell, they'll buy more. So it's very difficult if you're a company that turns through, that has a lot of new products and really counts on those new products to drive a lot of your performance. Being a vendor is really hard. 

Paul Sonneveld
Yeah, no, absolutely. All right. We're well over. It's 33 minutes, so I'm going to close us out. Thank you particularly to you, Jackson, for giving us your great questions. That was wonderful. But yeah, everyone, that is a wrap on today's episode of Marketplace Masters. Nathan, huge thanks to you. Really appreciate you really cutting through some of the noise around Amazon vendor forecasting and demand planning, and really sharing some great tips and know-how. Really helpful. I've written down SKU signs to have a look at as well. Always keen to see how people do this. So yeah, thank you very much. 

Now, if you joined us live, thank you for being part of this conversation. And if you've missed any part of today's discussion, don't worry, the full recording will be available shortly. So you can catch up and share it with your team afterwards. Now, before we sign off, just a quick reminder, Luke and I will be at Amazon Unboxed in Nashville in early November. And of course, don't forget to register for our Amazon Vendor Summit coming up in December. It's a great opportunity to pick up a whole lot of learning during the course of this two-day event designed specifically for Amazon Vendors. All right. With that in mind, my name is Paul Sonneveld. This was Marketplace Masters, and have a wonderful week. Take care.