I remember standing in the back of our primary fulfillment center in mid-January, staring at a literal wall of cardboard. We’d just survived a 5.3x return spike during the BFCM rush, and our floor space was physically running out. Every square foot was occupied by "zombie stock"—items that were technically sold but now lived in a purgatory of uninspected returns. It’s the ultimate nightmare for any DTC operator: having plenty of paper profit but zero liquidity because your cash is rotting on the shelves. I realized then that our outbound game was strong, but our demand forecasting for the reverse leg was non-existent. If you aren't obsessing over how many units are coming back just as much as how many are going out, you aren't running a business; you’re running a very expensive storage unit.
What is Demand Forecasting and Why Does It Kill Growth?
If you’re new to the logistics space, you’re likely asking, what is demand forecasting in the context of a 2026 supply chain? Simply put, it is the process of estimating future sales so you can make informed decisions about inventory, staffing, and marketing spend. It is the pulse of your brand. A faulty prediction doesn't just mean a few lost sales; it means a "death spiral" of trapped capital and warehouse bottlenecks.
Here’s where ops breaks: most brands only forecast for the "buy." They look at their Shopify dashboards, see a 20% month-over-month growth, and order 20% more stock. But they ignore the "return." In my experience, if your sales grow by 20%, your returns might grow by 40% if you hit a quality snag or a sizing issue. I recall an anecdote from a footwear brand in 2024 that over-ordered their spring line by 15,000 units because they didn't account for a 12% return rate on their winter boots. They had the boots, but they had nowhere to put the new sneakers. (Honestly, staring at a warehouse full of insulated boots in May is a special kind of stress).
Now the logistics math that matters: every pallet of stagnant stock in a 3PL like ShipBob costs you between $15 and $40 per month. If your forecasting and demand planning is off by just 100 pallets, you’re lighting $48,000 a year on fire. That’s a marketing manager’s salary or a whole new product development cycle gone because your demand for forecasting didn't include the reality of physical space.
Supply Chain Forecasting: The Art of the Demand Forecaster
To be a successful demand forecaster, you have to stop looking at spreadsheets in a vacuum and start looking at the "Full Loop." This is where supply chain forecasting gets complicated. You aren't just predicting what people want; you’re predicting the capacity of your manufacturers, the speed of your freight forwarders, and the efficiency of your 3PL labor.
But here’s what most Ops Managers miss: the "Reverse Logistics Lag." When an item is returned, it is "dead" until it’s inspected and restocked. If your demand and supply forecasting doesn't account for that 14-day window where the item is sitting in a UPS or FedEx truck, you will under-buy your hero SKUs. I’ve seen honest failure cases where brands had a $0 bank balance because their cash was tied up in 5,000 "Returns-in-Progress" that Narvar or Loop had marked as "received" but the warehouse hadn't actually touched.
So, what is a demand forecaster supposed to do when the data is messy? You build "buffers." But those buffers cost money. I’m of the opinion that the traditional warehouse model is actually the biggest obstacle to accurate demand forecasting models. When everything has to go back to one central hub, the data delay is too long to be actionable.
How to Forecast Demand in Supply Chain Using Modern Models
If you want to know how to forecast demand in supply chain settings, you need to move beyond simple moving averages. You need to look at demand forecasting models that include:
-
Qualitative Data: Feedback from your customer service team about why people are returning items.
-
Time Series Analysis: Seasonal trends (the "Post-BFCM Hangover" is a real data point).
-
Casual Models: How a 10% increase in ad spend impacts return volume three weeks later.
Now the logistics math that matters: a $27 return processing cost for a $19 resale item is a losing game. If your demand forecasting tells you that a specific SKU has a 30% return rate, your "Expected Profit" on that item needs to be adjusted down by at least 15% to account for the labor. We’ve seen brands use enterprise tools like Optoro or Happy Returns to manage these flows, but even the best software can't fix a broken physical route.
And let's be real—sometimes supply chain forecasting is just an educated guess. I’m still uncertain about how the rise of AI in "generative commerce" will impact these models, but I do know that the faster you get your inventory back, the less your forecast has to be "perfect."
How Closo Predicts Demand and Works with Returns
This is where the conversation changes. This is how Closo predicts demand differently. Instead of just looking at historical sales, we look at local density. We realize that if 40% of your sales are in the NYC metro area, 40% of your returns will be there too.
How Closo works with returns is by decentralizing the "Source of Truth." Instead of shipping that NYC return back to your warehouse in Ohio, we route it locally. We route eligible returns locally instead of sending everything back to the warehouse — cutting return cost from ~$35 to ~$5 and speeding refunds.
By utilizing return hubs, we provide the demand forecaster with real-time data on local inventory availability. This means you don't have to ship a "new" unit from Ohio to a New Jersey customer if a "restocked" unit is already sitting in a Manhattan hub. This isn't just logistics; it's a better way to Closo your inventory gaps. This level of agility turns your "Return Pile" into a "Live Shelf."
Comparison: Centralized Warehouse vs. Localized Hub Routing
Operators always ask me... "How do I start demand forecasting without a huge team?"
Common question I see: "I'm a founder doing $2M GMV. Do I really need complex demand forecasting models?" The short answer: No. But you do need to track your "Sell-Through Rate" vs your "Return Rate" at the SKU level. If you see a SKU that has high sales but an increasing return rate, that is a red flag for your next manufacturing run. (Parenthetically, I’ve seen more brands go bankrupt from "bad returns" than from "low sales").
And here is where most people fail: they treat every return as "bad luck" rather than a data point. Use tools like Loop or Happy Returns to ask why the item came back. If 50% of people say "too large," your demand for forecasting needs to adjust down for the Larger sizes and up for the Smalls in the next PO.
A question I hear from CFOs often... "Does Closo integrate with my ERP?"
CFOs always ask me about the "single source of truth." They worry that having inventory in local hubs will make the books messy. But how Closo predicts demand is by staying in sync with your tech stack. We plug into enterprise tools like ShipBob and NetSuite to ensure that the "Virtual Warehouse" is always accurate.
Now the logistics math that matters: if your "refund delay impact" is costing you 2% in credit card chargebacks and customer service labor, moving to an instant-refund local model saves you more than just shipping costs. It saves your brand's reputation. For more on how we bridge this gap, check out our brand hub for deep dives into tech integration.
Honest Failure: The Slow Refund and the Backlog Trap
I recall an honest failure case with an apparel brand in 2024. They had a great demand forecaster on staff who predicted a massive holiday season. They were 100% right. Sales were up 300%. But they didn't forecast the "processing labor."
Because they had a warehouse backlog, it took them 28 days to process returns. Customers got angry. They started filing disputes with their banks. The brand got hit with "Refund Delay" penalties and lost their "Top Rated" status on Google Shopping. They had the right demand and supply forecasting for the outbound, but they were blind to the inbound. By the time they cleared the backlog, they realized they had $100,000 in inventory that was now "off-season" and had to be liquidated at a 60% loss via Optoro. (The lesson: your forecast is only as good as your ability to process the result).
Conclusion: Balancing the Art and the Atoms
In the end, demand forecasting is about more than just numbers on a screen; it is about the physical reality of moving atoms in space. In 2026, you cannot afford to have your capital trapped in a "centralized" bottleneck. While the art of being a demand forecaster will always involve some uncertainty, the goal is to shorten the distance between the data point and the decision. Decentralized logistics—keeping the inventory near the customer—is the only way to stay agile in a market that moves at the speed of a viral tweet.
We route eligible returns locally instead of sending everything back to the warehouse — cutting return cost from ~$35 to ~$5 and speeding refunds. Would you like me to run a "Return Density Analysis" on your last 90 days of sales to see how much cash you could unlock with a localized return hub
