You sell on Shopify. You sell wholesale. You sell on Amazon. Each channel has its own dashboard, its own velocity numbers, its own seasonal quirks. So naturally, someone on your team built three separate spreadsheets to forecast each one. Maybe they're even color-coded. It looks organized. It's not.
Here's the problem: those three spreadsheets are all pulling against the same physical SKU pool. And the moment they each recommend a reorder independently, you're not tripling your coverage. You're setting yourself up for over-ordered inventory, tied-up cash, and a warehouse full of units that all three channels simultaneously decided to stock up on.
We've seen this exact scenario play out at a mid-market apparel brand that was running Shopify DTC, two wholesale accounts, and an Amazon Seller Central operation in parallel. Their Q4 demand plan looked bulletproof on paper. Each channel showed healthy sell-through projections. They ordered to cover all three. In January, they had 14 weeks of stock across 23 SKUs and a $380,000 inventory write-down to explain to the board.
Why Separate Forecasts Break at the SKU Level
The logic of channel-by-channel forecasting seems sensible. Each channel behaves differently. Amazon has different seasonality than wholesale. A DTC promo spike on Shopify doesn't affect your net-30 wholesale buyer. These are real differences worth modeling.
But here's what separate forecasts don't account for: demand isn't additive across channels when supply is shared. If your Shopify model recommends ordering 500 units and your Amazon model independently recommends 400 units of the same SKU, you don't need 900 units. You need to know how those 900 units of projected demand will actually distribute across channels, in what sequence, and with what priority logic.
Separate spreadsheets have no mechanism for that. They each assume the supply pool is infinite. It isn't. The result is double-counting. Both models effectively count the same projected customer purchases against separate demand buckets, but there's only one warehouse. You order to satisfy both forecasts independently. You end up with stock that was counted twice.
Double-counting is the silent killer of multi-channel inventory operations. It doesn't show up as a forecasting error. It shows up as a cash flow problem three months later.
How Channel Weights Actually Work in Consolidated Models
A consolidated demand model doesn't average your channel forecasts. That's a common misconception. It weights them.
Channel weighting is the process of assigning each channel a relative influence on the consolidated demand signal based on factors like data reliability, historical accuracy, lead time sensitivity, and revenue priority. In practice, this means your Amazon channel might carry a 40% weight, your Shopify DTC a 35% weight, and your wholesale channel a 25% weight, depending on how much of your total revenue each drives and how predictable each channel's velocity has been historically.
The key insight is that these weights are not fixed. They should adjust based on confidence. A wholesale channel with a committed purchase order from a single large buyer is extremely high-confidence demand. A single-buyer account generating 25% of your projected channel demand is, paradoxically, less reliable than a diffuse set of 300 independent Shopify customers, because losing that one account means losing the whole slice.
Practical note: the right weight for a channel isn't the one that reflects its revenue share. It's the one that reflects how much you trust its demand signal. A channel with lumpy, buyer-dependent demand needs a lower confidence weight even if it drives 30% of top-line revenue.
When weights are applied correctly, the consolidated forecast reflects one unified demand number per SKU, per planning period, with each channel's contribution adjusted for its reliability. That number drives a single replenishment decision against a single SKU pool. No double-counting. No competing models pulling in opposite directions.
The Amazon Problem: Velocity vs. Commitment
Amazon deserves its own section, because it creates a specific confidence weighting challenge that wholesale doesn't.
Amazon velocity data is high-volume and high-frequency. You can pull daily sell-through, conversion rates per ASIN, and traffic data in near-real-time. It looks like a very confident signal. The problem is that Amazon demand is reactive rather than committed. A wholesale buyer signs a purchase order. Amazon customers click. Those are fundamentally different levels of demand certainty.
In our experience, Amazon velocity data is best treated as a leading indicator rather than a hard demand commitment. That means its confidence weight in a consolidated model should reflect the volatility of that velocity, not just its volume. An ASIN with high daily sales but a 60-day sales rank history that shows 35% week-over-week variance is not a stable signal. It shouldn't carry the same confidence weight as a wholesale PO with a fixed ship date.
Compare that to a wholesale channel running on net-30 terms with a single large retailer. That buyer's PO is a committed number. But it's also a single data point. One buyer. One decision. Lose that account and your demand signal drops to zero for that channel. The confidence weight for that buyer should reflect the concentration risk, not just the purchase commitment.
Running these two very different demand types through the same undifferentiated forecast model, without confidence weighting, is how operators end up over-ordering into Amazon volatility while under-serving their wholesale commitments. Or the reverse. Both happen. Neither is acceptable.
What Configurable Confidence Weighting Changes
Configurable channel confidence weighting is the mechanism that makes a consolidated model usable in practice. Without it, you're stuck with either a rigid statistical model that treats all demand equally, or a fully manual override process where someone is adjusting weights in a spreadsheet. Both fail under real operating conditions.
What configurable weighting does is let you encode operator judgment into the model systematically. You can set a floor weight for wholesale that accounts for contractual commitments regardless of velocity trends. You can set a dampening factor for Amazon that reduces its effective weight during periods of high velocity variance. You can define a confidence decay rate that automatically reduces the weight of stale wholesale orders past a certain lead time threshold.
These aren't magic settings. They require calibration. In our tracking across mid-market operators who've moved from siloed forecasts to consolidated models, initial weight configurations typically need three to four planning cycles of adjustment before the consolidated forecast outperforms the best individual channel forecast. That's 90 to 120 days of tuning. Not instant. But the compounding accuracy gain after that calibration window is real.
The operators who skip confidence weighting and go straight to a simple blended average see a different outcome: the consolidated number looks smoother and more defensible in a planning meeting, but it's just as wrong as the individual channel forecasts. Averaging noise doesn't remove it. Weighting it does.
Closing the Loop: From Consolidated Forecast to Replenishment Decision
The consolidated demand number is the input, not the output. The output is a single replenishment decision per SKU that accounts for lead times, safety stock logic, and channel fulfillment priority.
Channel fulfillment priority matters more than most operators realize. If you're managing shared inventory across Shopify, wholesale, and FBA prep, you need explicit rules for how to allocate available stock when total demand across channels exceeds supply. Without those rules, the team defaults to fulfilling whoever shouts loudest, which usually means wholesale commitments get honored and Shopify orders suffer stockouts, or FBA inventory gets drained for a promo and wholesale can't be fulfilled.
A consolidated forecast model doesn't solve the allocation problem automatically. But it surfaces it clearly. Instead of discovering three weeks before a wholesale ship date that you over-allocated to Amazon FBA and can't fulfill the PO, you see the constraint at planning time. That's the difference between a reactive scramble and a deliberate tradeoff.
Honest take: most multi-channel inventory problems aren't forecasting problems. They're visibility problems. The demand was always there to model correctly. The issue was running three separate models that couldn't see each other. Consolidation fixes the visibility. The accuracy follows.
Key Takeaways
- Separate per-channel forecasts applied to a shared SKU pool cause double-counting, not better coverage. Both models assume infinite supply.
- Channel weights in a consolidated model reflect demand signal reliability, not just revenue share. A committed wholesale PO and 500 daily Amazon clicks are not equally confident inputs.
- Amazon velocity is a leading indicator, not a committed demand signal. Weight it for variance, not just volume.
- Single-buyer wholesale channels carry concentration risk. High commitment, high fragility. Both factors belong in the confidence weight.
- Configurable weighting needs a calibration window, typically 90 to 120 days, before it consistently outperforms individual channel forecasts.
- Consolidated forecasting surfaces allocation conflicts at planning time instead of fulfillment time. That's the operational advantage.
If your team is still maintaining three separate forecast spreadsheets for a shared SKU pool, the first step isn't better modeling. It's consolidating the demand view. Everything else follows from that. Talk to us about how Stockorlo handles multi-channel demand consolidation out of the box.
