November 30, 2024. 3:30 in the morning. The day after Black Friday.
I'm sitting in the office at our Qianhai warehouse, watching the night shift through the glass, when our customer service lead calls. Again. "Chen, that Bluetooth earbud listing shows 800+ units available on Amazon. But across our three warehouses, we can actually ship 540. We've already got 320 orders we can't fulfill. What do we do?"
Let me be straight with you: the scariest thing in warehouse management isn't running out of stock. It's the system saying you have stock when you don't. That night, our CS team apologized until 7 a.m., the night crew worked back-to-back shifts, and next-day picking efficiency fell off a cliff. Overselling doesn't start in the warehouse — it starts in the numbers nobody agreed on.
Where Those 320 Oversold Units Came From
After the dust settled, I pulled every record apart. The system showed 860 sellable units: 400 at Qianhai, 300 at Dongguan, 160 at Zhongshan. Sales planned their promos around 860. But buried inside that 860: 120 units at Qianhai were sitting in the QC hold area, returns not yet re-inspected. 90 units at Dongguan were reserved for Walmart orders. 110 units at Zhongshan were already on pick tickets, sitting in an open wave. 120 + 90 + 110 = 320. Not one unit more, not one less.
Each warehouse's books balanced. But three balanced books don't add up to one usable number. Until you subtract in-transit, in-pick, QC-hold, and reserved units, the number on the screen is decoration.
We got burned by "in-transit" again in June 2025. A container of 800 units showed as "arrived" in the system. In reality, the box was still sitting at Yantian Port, uncleared. Sales launched a flash sale and moved 1,200 units overnight. The container arrived three days late. We covered the 400-unit gap with emergency transfers and burned several thousand dollars in extra freight. If it hasn't crossed the dock door, it doesn't count as sellable. Period.
The Dead Ends We Tried First
Our first attempt was real-time sync. All three warehouses pushed numbers to HQ every minute. The big screen looked alive — numbers jumping, sales loved it. Then the night picking peak hit: waves opening and closing, numbers bouncing up and down. CS saw "50 available" and locked the listing; five minutes later it read "200 available" and sales came screaming at me. Real-time dashboards have one fatal flaw: they treat every mid-process fluctuation as fact.
So we swung the other way — manual spreadsheets. Every morning at 9, each warehouse emailed me an Excel file and I consolidated them. Accurate, sure. But in a peak season where we can sell three days' worth of stock in one day, the numbers were yesterday's news by the time the sheet was done. CS kept taking orders off stale data. Same overselling, slower.
What We Actually Do Now
Today there's a big screen on the office wall. Three warehouses, the CS lead, and the sales ops team all look at the same board. Here's what it shows:
| Field | Why it's there |
|---|---|
| Product / SKU | Only bestsellers and A-class items — you can't watch everything |
| Available per warehouse | One column per site, accountability per site |
| In-transit / In-pick / QC / Reserved | The four deductions, each in its own column — you can see exactly where stock is stuck |
| Total available | System balance minus the four deductions, minus a safety buffer |
| Daily sales velocity | 14-day average, 7-day in peak season |
| Days of cover | Total available ÷ daily velocity — the one number that matters |
| Alert light | Green / yellow / red, readable at a glance |
| Last updated | Down to the minute |
The formula is taped to every warehouse supervisor's monitor: Available = system balance − unpicked wave units − QC hold − platform reservations − in-transit not yet received − safety buffer. Our safety buffer is one day of sales; two days for hero SKUs. The board refreshes every 15 minutes — 5 was too twitchy, one hour too slow. Warehouse screens stay on all day, the CS lead has it on her phone, sales ops check the red/yellow lights each morning, and the night manager screenshots it into the work group hourly. When something goes wrong, the screenshot is the evidence.
Here's the thing nobody tells you: when we first launched the board, sales ignored it completely. Their exact words: "Your warehouse numbers — have they ever been right?" I couldn't argue. They hadn't been. The turning point came in August 2025: The board showed 2 days of cover. Sales didn't believe it, ran the ads anyway, and stocked out on day three — over ten thousand dollars of ad spend down the drain. After that, sales checked the board before me every morning. A dashboard earns trust not through meetings, but by being right once when it counts.
One more detail: the "owner" column lists the on-duty supervisor's name at each warehouse. When numbers don't add up, you go to the name first, then the cause. Put a person's name on the wall and the numbers get serious fast. We also made one hard rule: if you think the board is wrong, don't complain in the group chat — go count the physical stock first, then talk. Every time someone counted, the board was right.
Oversell Alerts: Thresholds and the Action Checklist
We keep exactly two alert levels. Simple is the only thing that gets followed:
- Yellow: days of cover ≤ 3. CS gets notified and preps their scripts; sales pause new promo submissions.
- Red: days of cover ≤ 1. Auto-lock — ads and lightning deals for that SKU stop, and CS stops promising ship times.
The post-alert action checklist is posted on the CS and sales walls, with accountability if deadlines slip:
- Within 10 minutes: CS locks the listing — pause off-site ads and deal scheduling for that SKU.
- Within 30 minutes: the warehouse confirms physical counts across all three sites and replies with the "real shippable number." The warehouse number is final.
- Within 1 hour: sales picks one of three — transfer stock from the nearest warehouse, delist temporarily, or switch to pre-order mode.
- Within 24 hours: the replenishment PO goes out, and the factory/supplier lead time gets written into the board.
What One Oversell Really Costs
A lot of owners think overselling just means "cancel an order, eat the shipping." Let me give you the real math:
| Cost item | The damage |
|---|---|
| Platform penalties | A cancelled order costs tens of dollars in fees — but the real hit is a defect-rate breach that throttles your listing (check each platform's current policy) |
| CS labor | At least 15 minutes per order: apologize, modify, soothe. In peak season a rep handles maybe 30 a day |
| Negative reviews | One "seller cancelled my order" review and that listing is dead for two weeks — the biggest invisible cost |
| Intercept / reroute | Stopping a parcel mid-shipment costs a few to a dozen dollars per package, and it doesn't always work |
| Team morale | The night crew works doubles, next-day picking efficiency drops ~20%, and mispicks climb with it |
Honestly, the accounting losses are small change. Reviews and throttling are what kill you. You nurse a hero listing for three months, and one big oversell sends you back to zero.
The Counterintuitive Part
People ask me: "Chen, now that the board is live, overselling is gone, right?" Not exactly. In the first three months, our oversell incidents actually went up. Why? Sales could suddenly see the combined available number across all three warehouses — a bigger number — and got bolder. Promos they used to skip, they started taking.
That's the trade-off nobody warns you about: transparency makes sales more aggressive; set alerts too tight and you lock listings every day until nobody takes the alerts seriously — cry wolf three times, and the fourth real oversell gets ignored. So we deliberately set yellow wide and only let red bite. Whether an alert system works isn't just a technical question — it's a human one.
If I had one thing to tell fellow warehouse folks: the board itself is the easy part. The hard part is getting every department to accept the board after the definitions are unified. In our warehouse now, when anyone questions a number, we check the screen, then the screenshot, then the physical stock — by the end of those three steps, the argument is usually over.
By Chen Feng, Warehouse Operations Manager
