We Forecast $1.2M. We Did $2.3M. In One Hour.
The forecast for a client brand's first Prime Day on Amazon was $1.2 million for the entire day. That number came from conversion rate estimates, expected traffic multipliers, and the team's best read on how aggressive Prime Day demand would actually be for premium cookware.
In the first hour, we were pacing $2.3 million.
That's not a success story dressed up as a challenge. Being right by that much creates its own set of problems, and those problems are the point.
What happens when you over-perform by 90%
The most immediate problem wasn't running out of stock. It was getting enough stock into Amazon in the first place, which we very nearly couldn't do.
A client brand runs seven warehouses globally: US facilities in California, Texas, Pennsylvania, and Illinois; a Canadian facility in Colton and another in Ontario; plus UK and Australia. FBA inventory has to come from somewhere, and "from somewhere" matters a lot at scale.
Amazon imposes supply limits. For a client brand, with large oversized items like the countertop oven and multi-cookers, those limits were tight. Amazon was capping individual shipment sends at roughly 5,000 units. The warehouse couldn't process more than 100 to 150 pallets in a single inbound. Carrier detention fees were accruing because Amazon's receiving team was slow even before Prime Day started.
We needed to get 10,000 units into FBA. We had to split it into two consecutive weeks just to work around the limit walls. That only works if you start early enough.
The California-first rule
Amazon's routing algorithm told us to split inventory across Texas, Pennsylvania, and Illinois fulfillment centers. Amazon's system looks at where FCs have capacity and routes accordingly. It doesn't know anything about your warehouse infrastructure.
We pulled from California only.
A client brand's California warehouses, CAO2 and CAO3, are positioned closest to Amazon's major West Coast FCs. Freight cost per unit is lower. Transit time is faster, which means FBA check-in happens before Prime Day rather than during it, which means stock shows as available when customers are actually buying. The other warehouses have their uses. For US FBA during peak, California is the right answer regardless of what Amazon's algorithm recommends.
This became a hard rule. Any time Amazon's routing pointed east, we overrode it.
Planning for a number you can't predict
You can't predict Prime Day demand with precision. The $1.2M number was a careful forecast, built from the data we had. But Prime Day creates a step-change in traffic that no model captures cleanly, because nothing in normal operating history generates a comparable signal.
What you can do is plan for the operational floor and the operational ceiling separately.
The floor: have enough inventory in FBA to cover your best-case forecast. The ceiling question is different. It's: if demand comes in at 2x or 3x what we expect, what breaks? For a client brand, the answers were supply limits (can we send more?), warehouse throughput (can we process the outbound?), and Amazon receiving speed (will it arrive in time to count?).
We built the shipment plan around the ceiling scenario. That meant starting shipments earlier than necessary, using the 50/50 split (send half this week, half next week) to keep FBA flowing without hitting limit walls, and filing cases to negotiate supply limit increases before Prime Day, not after. Detention charges from Amazon's slow receiving went into a case file too. Amazon reimburses documented detention. Most brands absorb that cost because the process feels unclear. It isn't. You need timestamps, appointment confirmations, and delivery records.
What broke anyway
Australia. Australia FBA was, in the words of the team at the time, a mess. We couldn't get shipments created properly. The account setup had structural issues that took months to resolve. International expansion always lags the US buildout, but in this case it lagged enough that Australia was a non-participant in Prime Day while the US channel was hitting numbers nobody planned for.
UK had its own issue: standard-size storage capacity was exhausted. The fix was to pivot UK FBA to oversize-only: large countertop ovens and multi-cookers only, with standard-size held at the warehouse and fulfilled via FBM until storage freed up. It worked, but it was reactive. Had we run the ceiling-scenario analysis for UK earlier, we'd have restructured the storage allocation weeks before it became a problem.
Why over-preparation was the right call
The cost of being ready and not needing it: extra freight spend, a few pallets that cycle back from FBA. The cost of being right by 90% and not being ready: a stockout when demand is at its annual peak, rank loss that takes weeks to recover, and missed sales that don't come back.
Those are not equivalent outcomes.
The supply limit audit needs to happen two weeks before Prime Day, minimum. Late discovery means missed sales, because you can't move stock through FBA fast enough no matter how much sits in the warehouse. Amazon's defaults for routing, receiving, and limit-setting were wrong for a client brand's situation. The team that built its own rules around those constraints did well. The ones that just follow Amazon's defaults mostly don't.
$2.3 million in the first hour is a good problem. It's still a problem.
Want this run on your catalog?
This is the kind of problem we work every week inside live accounts. Bring your stock report to a 30-minute Fit Call, we’ll tell you which SKUs are at risk in the next 90 days. If we’re not the right fit, we’ll say so.