Benchmark, correct as of August 2026
Ten of our deployments, published as data.
Sentispec Inventory runs across many locations globally. These are ten of those sites, in five countries, from 5,000 to 85,000 pallet locations, with accuracy after go live running from 99%+ to 99.99%+. They are published because the figures were measured and the customers allow them to be shared at location and sector level. Every column has its measurement method stated below, including what the figures do not control for.
The ten sites
Every live deployment, in full.
Ten rows, no selection: every site Sentispec Inventory has taken live and is able to publish.
| Location | Pallet locations | Accuracy after go live | Sector | Value realised | Break even |
|---|---|---|---|---|---|
| Duisburg, Germany | 30,000 | 99.9% | Retail and fashion | 95% lower stocktaking cost, 90% fewer loading errors | 1 month |
| Cuxhaven, Germany | 5,000 | 99%+ | Manufacturing | 95% fewer stock errors, production stops eliminated at €7,000 loss per minute | 1 month |
| Łódź, Poland | 15,000 | 99.99% | Manufacturing, retail and fashion | 90% lower stocktaking cost, 90% fewer loading errors | 3 months |
| Bremen, Germany | 35,000 | 99.5% | Retail and fashion | 85% lower stocktaking cost, 90% fewer loading errors | 3 months |
| Mechelen, Belgium | 65,000 | 99.99%+ | Retail | 95% fewer put away errors, 90% lower stocktaking cost | 3 months |
| Venlo and Roermond, Netherlands | 85,000 combined | 99%+ | Retail and fashion | 95% lower stocktaking cost | 4 months |
| Willebroek, Belgium | 29,000 | 99.99% | Retail and fashion | Contract renewed, 90% lower stocktaking cost, 9 point accuracy gain | 5 months |
| Kolding, Denmark | 25,000 | 99%+ | Food | 80% lower stocktaking cost, 80% fewer loading errors | 6 months |
| Horsens, Denmark | 40,000 | 99%+ | Retail | 85% lower stocktaking cost, 90% fewer loading errors | 6 months |
| Geel, Belgium | 7,000 | 99.99%+ | Pharma | 85% lower stocktaking cost, alignment with safety regulation | 6+ months |
Two sites sit inside published case studies, with the full operating detail rather than a row of figures: Cuxhaven, a wind turbine manufacturer protecting €7,000 a minute of uptime, and Geel, a pharmaceutical cold store scanned in under four hours.
Which system each site runs. Across the customer base there are live deployments exchanging stock data and running discrepancy analysis on SAP EWM, Manhattan, Blue Yonder, Körber, Infor WMS and Astro WMS, and on sites running no WMS at all. Which of the ten sites above runs which system is not published, at our customers' request. It makes no difference to the figures: the reconciliation takes a stock file, and the mechanism is the same behind all of them.
Method
What was measured, when, and how.
How each column in the table above was measured.
- Accuracy after go live. Location level accuracy: whether the pallet recorded in a position matches the WMS, including positions the WMS expects to be empty. It is measured against a controlled count run at go live, then monitored through ongoing counts afterwards. It is not a count of stock value or of what is inside a carton.
- Pallet locations. The number of racking positions covered by the deployment, stated by the customer and used to size the site. It sets the scale of the count, not the count frequency.
- Value realised. The outcomes each customer reports against their own pre-deployment baseline: stocktaking cost, loading errors, put away errors or stock errors, stated as the percentage change that customer measured.
- Break even. The point at which the value realised at that site, measured in euro against that site's own prior cost of stocktaking labour and lost pallet search, exceeds the cost of the setup fee plus the pro-rated annual fee for that site. The arithmetic behind that calculation is published in full on the pricing page.
What this benchmark does not control for. These are operational deployments, not a blinded or randomised study. Sites differ in size (5,000 to 85,000 pallet locations), sector, starting accuracy and local labour cost. None of them ran a parallel head to head count against the manual method at the same time in the same building. Value realised and break even are calculated from each customer's own cost model, agreed with that customer, not audited by an independent third party. They are ten separate site results, not a controlled trial.
Counting methods, compared
What each method asks of a warehouse.
Accuracy and speed get the headlines. What decides which method fits a building is capital cost, deployment time, whether anything about the facility has to change, and who has to be there to run it.
| Method | Capital cost | Deployment time | Facility change | Cold store suitability | New vehicle in a working aisle | Who operates it | Driver needed |
|---|---|---|---|---|---|---|---|
| Manual, walked count | None | None, already in use | None | Poor. Labour cost and regulatory constraints run higher in cold storage, and high locations need a scissor lift | No | Warehouse staff, on foot | No |
| Hand scanner | Low, scanners typically already owned | None, already in use | None | Poor, for the same reasons as a manual count | No | Warehouse staff, on foot | No |
| Inventory drone | Vendor hardware or subscription; pricing not published | Not published by the vendor | Housing and charging infrastructure for the vehicle | Not independently measured by Sentispec | Yes, airborne | Runs autonomously; no continuous operator needed during a count | No |
| Autonomous ground robot | Vendor hardware or subscription; pricing not published | Under two weeks, Dexory's own published claim | Housing and charging infrastructure for the vehicle | Not independently measured by Sentispec | Yes, self driving | Runs autonomously; no continuous operator needed during a count | No |
| Camera on a forklift (Sentispec) | None. The unit is leased inside the annual fee | Five days | None. Nothing fitted to vehicles, nothing changed in racking | Suited. A cold store site scans down to minus 23 degrees, one employee in under four hours | No | The customer's own forklift driver, during normal shifts | Yes, for the duration of a run |
Dexory, a maker of autonomous scanning robots, publishes a claimed deployment time of under two weeks and a claimed ROI of under six months. Verity, a maker of autonomous indoor drones, publishes a claim of 99.9% accuracy and near zero churn across more than 200 warehouses. Both are those vendors' own published claims, not measurements Sentispec has taken.
What Sentispec has not measured independently. Sentispec has not run a controlled trial against a competitor's hardware in the same warehouse, and has no independent measurement of drone or ground robot accuracy, deployment time or cold store performance. Every competitor figure above is attributed to that vendor's own published material.
A driver and a forklift for the length of a run is a real cost against a vehicle that counts unsupervised. What it buys is a five day deployment, no capital purchase and no facility change. The two comparisons are on inventory drones and autonomous ground robots. To run the figures against your own count cost, use the ROI calculator.
Next step
See the recording behind any row in the table.
A 30 minute demo using a recording from a site with your racking and your label format, so you can judge the method for yourself.