Every year, the same call comes in around Thanksgiving: "We need a full physical count next week — can you help us plan it?" And every year I give the same answer: you're about two months late.
I learned this the hard way early in my career, watching a distribution center in Fontana try to do a full wall-to-wall count the week before Christmas. Temp workers counted the same pallet twice, nobody knew how to handle damaged goods, and the auditors ended up expanding their testing so much that the whole exercise cost roughly three times what a properly planned count would have. Since then I've helped plan year-end counts for dozens of operations, and the pattern is always the same: the count itself isn't the problem. The planning is.
There are really only three ways to do a year-end inventory count. Picking the wrong one costs more than picking none.
Know what you're actually buying with each method
A full physical count is the blunt instrument. You shut down operations — or at least freeze all receiving and shipping — and count every slot, every pallet, every SKU. It's the one result auditors never argue with. The price is downtime. For a busy distribution center, one lost day of throughput, plus labor and late-shipment penalties, can wipe out any money you saved by skipping a count earlier in the year. And here's the thing people forget: the post-2020 peak season keeps stretching. The "quiet week between Christmas and New Year's" that full counts used to hide in is getting smaller every year.
Sample counts speak the auditor's language, not the warehouse's. Your auditor picks a percentage of SKUs or locations, counts them, and if the results reconcile with your books, they accept the whole population. But there's a catch that nobody tells you upfront: sampling only works if your day-to-day records are already solid. I've seen an operation in the Inland Empire fail so badly on the sampled items that the auditors expanded the count to the entire building. That year's count cost multiples of what a planned full count would have run.
Cycle counting spreads the work across the whole year. You count a slice of inventory every day or every week, so by December the entire warehouse has been counted several times over and your book-to-physical accuracy stays high all year. The best-run warehouses I know barely do anything special at year end — they hand the auditor twelve months of cycle count records and a small validation sample, and the auditor signs off. The catch: cycle counting needs real planning. You need a WMS that can manage count tasks, trained people executing them, and an ABC classification done right from day one. If you haven't started by October, it won't save this year-end.
The four questions that make the decision for you
Stop asking which method is "best." Each works fine; the question is which one fits your operation. Here's the checklist I use with every warehouse manager.
First: your SKU profile and ABC distribution. A few thousand SKUs with the A-items concentrated in a handful of part numbers is easy to sample or cycle count. A fifty-thousand-SKU e-commerce operation with a long tail is a full count nightmare — cycle counting is the only sane path. Second: what does your auditor actually require? Call them now, in October, not in December. Some auditors accept a year's worth of cycle count records plus a year-end validation sample; others insist on one point-in-time count. Get that answer in writing this month. Third: your December labor reality. You know exactly what quality of temp workers shows up in December. Counting is detail work, and temp error rates run multiples of your regular crew. A full count executed entirely by temps can produce numbers worse than no count at all. Fourth: can your WMS actually support this? Cycle counting without system support — task assignment, recount workflows, freeze management — is just wandering around with a clipboard. If your system can't do it, don't pretend.
My honest take, after two decades in this business: for most operations, cycle counting is the long-term answer, but it cannot save this year's year-end. If you've been cycle counting since spring, do a small validation sample in December and your auditor will likely accept it. If you haven't done any systematic counting this year, stop dreaming and plan a full count — but run a pre-count first to clear dead stock, damaged goods, and the ghost inventory sitting on your books. Otherwise December will be a very long week.
Two more traps worth flagging early. The first is cutoff. At the moment of the count, who owns the freight in transit, the goods sitting in QA hold, and the shipments that left the dock but haven't posted yet? That definition has to be agreed with finance and the auditors before count day, in writing. I've seen a third of all count variances trace back to cutoff misalignment — the counting was perfect and the numbers were still wrong. The second is blind versus open counts. Open counts hand counters the book quantities, which is fast but invites people to "count to the book" and bury variances. Blind counts give only locations, which is accurate but slow and rework-heavy. My practice: blind counts for A-items, open counts for low-value C-tail, B-items depending on staffing. Don't have this debate for the first time on count day.
Here's your timeline. October, right now: pick your method, align with your auditor in writing, decide whether you're shutting down or staying live, and start clearing dead and damaged stock. November: run a pilot or pre-count, walk through the full process once, train your counting crew, and print backup count sheets and location labels. December: execute, double-check every location with a large variance using a second pair of eyes, and produce the variance report within 24 hours of finishing — investigate the differences, don't just adjust the books. Tape this to your office wall, and the year-end count stops being a horror story.
