CUSTOMER DISCOVERY · EVIDENCE REVIEWED 8 SEPTEMBER 2026

A real problem. A narrower product hypothesis.

Audience: inventory-control leads resuming an isolated transfer discrepancy, with reliable existing counts and a small affected zone. Countback proposes the next measurement before the lead approves a reconciliation.

What operators already report

CHS reports that discrepancies forced recounts and delayed reopening warehouses; its own digital counting app reduced its inventory process time. This is first-party evidence that recount and reconciliation effort matters. CHS is not a Countback customer, and its results are not Countback results. CHS, 27 March 2024.

Microsoft's warehouse documentation explicitly supports blind counts, repeat attempts and supervisor review of differences. It also says open cycle-count work does not necessarily block stock movements. Countback therefore fits only a deliberately isolated investigation; it cannot safely assume an ordinary live warehouse is frozen. Microsoft: cycle counting.

Odoo already records count differences and an adjustment's movement history. Existing software is a serious baseline. Countback's proposed addition is choosing a diagnostic measurement from earlier evidence, not replacing the stock ledger. Odoo: inventory adjustments.

Named workflow context: Cheekky

The team supplied Cheekky context: weekly counts, garments moving through the work room, completed shelf and dispatch bags, and inventory discrepancies becoming loss when unresolved. This is supplied workflow context, not an independently observed Countback trial. Constant movement may violate the prototype’s isolation assumption; that is a validation question, not evidence that the model already fits. Read the exact source boundary · Public business identity.

The buying decision

Use Countback only when the avoided recheck effort exceeds all incremental work: preparing the case, verifying scope, collecting any extra historical count, coordinating the freeze and reviewing the result. If the same complete archive is already available to the same solver, retaining it in another format gives the same counting advantage.

The public calculator evaluates a selected synthetic path using your own time assumptions. It is a break-even test, not a measured return or an estimate of how often your operation will benefit.

What would validate the product?

1. An inventory lead supplies one de-identified, completed incident and confirms whether the supported premises actually held.
2. Record the existing process and its total labor, including investigation and review. Include historical evidence that was already available.
3. Replay Countback blind to the final quantities, revealing only requested counts. Separately record import, verification, freeze, count and review work.
4. Publish consented inputs, timing method, outputs, baseline and failures. Prefer a second operator and a second incident before generalizing.

Current status: no customer interview, Countback pilot, adoption or measured operational savings. Public pain evidence supports the audience hypothesis; it does not establish product-market fit. The synthetic benchmark is a mechanism evaluation.

Help falsify the hypothesis

Manage inventory investigations? Tell us what our model misses, how you handle rechecks today, and whether a completed incident fits. A negative result is useful. Do not upload confidential stock, customer or employee data.

Share an operator evaluation on GitHub ↗

No feedback is published on your behalf; you review and submit it yourself. Public issues are visible to everyone.