The robot is the sensor. The digital twin is the asset. Financial confidence is the product. See every shelf in your store, including the ones 23 feet up, and act on what you find the same day.
More than 80% of retail transactions still happen in physical stores, and most of those stores run on fragmented inventory data. The result is a gap between what the system says is on the shelf and what a shopper actually finds there.
of inventory records are inaccurate at any given moment
sales lift reported after a full store-wide inventory audit
of out-of-stock causes sit inside the store, in stocking and forecasting
And a quarterly count does not see it either. What is missing is not effort. It is a current, complete, trustworthy picture of the store.
Physical AI, edge computing, computer vision, RFID, Lidar, depth perception and built-in lighting, all in a single robot running on a single data model.
Not a robot, plus a fixed camera product, plus an RFID add-on, sold and priced separately and stitched together afterwards.
Multimodal by design means the shelf reading, the tag reading and the spatial measurement describe the same store at the same moment. That is what makes the output an asset instead of three reports that disagree.
Built-in lighting makes capture quality independent of store lighting. Zippedi works overnight, with the lights off and the aisles empty, which answers the shopper-perception question before anyone has to ask it.
A 7 metre camera reaches the top of a high-bay rack, so overhead stock is located, not estimated. That matters most in home improvement, DIY and club formats, where the difference between "out of stock" and "in the building, six metres above the aisle" is the difference between a lost sale and a pack-down instruction.
Gondola shelves are the easy part. Ask any platform how high it sees, in feet, and see whether it publishes a number.
Every service below runs off the same nightly scan. No extra hardware, no second visit, no separate subscription for the scan frequency that makes the data useful.
The list gets longer every year, and it gets longer because customers ask for the next one.
Zippedi has run continuous retail relationships for more than eight years, on four continents*. In that time the platform went from shelf reading to a services portfolio that includes GenAI presentation scoring, a virtual store walk and a supplier data service that turns store data into a revenue line for the retailer.
16 services from one scan, 5 of them in beta with customers today. The supplier data service turns your shelf data into a commercial asset you own.
More than 50 engineers work on the platform. Five services are in beta right now, each one started by a customer request.
Relationships measured in years, not pilot cycles.
Initial training takes about six weeks to produce usable data from a new customer. Once the models are trained for your formats, additional stores go live in three days each. Rollout speed is not the same thing as pilot speed, and rollout speed is where a chain-wide programme is won or lost.
Estimated figures from customer deployments. Results vary by format, store size, assortment and baseline data quality. These figures are indicative and do not constitute a guarantee of results.
| Format | Sales | Labour saved per store / month | ROI | OSA |
|---|---|---|---|---|
| Grocery / supermarket | +1 to 2.5% | 150 to 300 h | 3x to 7x | +4 to 8 pts |
| Home improvement / DIY | +0.75 to 2% | 200 to 350 h | 5x to 10x | +1.5 to 4 pts |
| Multi-department | +1.5 to 3% | 150 to 250 h | 4x to 7x | N/A |
Ranges derived from customer deployments. Results vary by format, store size, assortment and baseline data quality. These figures are indicative and do not constitute a guarantee of results.
Pending written authorization from Zippedi to publish this case study. Figures shown are the verified CAINZ results, not the older home page set.
* Figure pending written confirmation from Zippedi. See the validation list.
Whichever platform you choose, these are the questions that separate a demo from a deployment. Here is how Zippedi answers each one.
Zippedi's camera reaches 7 metres, about 23 feet, and locates overhead stock rather than inferring it.
Ask for a number in feet. Most platforms do not publish one, which usually means the answer is the top of a standard gondola.
No. Built-in lighting makes capture quality independent of store lighting, so Zippedi scans overnight with the lights off.
Ask what happens to image quality in a dim aisle, and whether scanning has to happen during trading hours with a human escort.
One device, one data model. Computer vision, RFID, Lidar, depth sensing and lighting ship together.
Ask whether the shelf reading, the tag reading and the spatial data come from one system or from separate products that have to be reconciled, and whether each one is priced separately.
You do. The robot is resident in your store and scans on your cadence, one to four times a day.
Ask whether the platform is resident or shared, whether scanning is tied to another route such as floor cleaning, and whether increasing the frequency changes what you pay.
You own your operational data.
Ask who holds the licence to it, for how long, whether that licence is transferable, and whether the same shelf data is also being sold to brands. The answer to that last question is not always in the contract you are shown first.
Zippedi has live customers on four continents*, across home improvement, DIY, grocery, multi-department and club formats.
Ask for deployments outside a single country, and outside a single store format.
Sixteen services from one scan, five of them in beta with customers today, including GenAI presentation scoring and a supplier data service.
Ask what shipped in the last twelve months, and who asked for it.
You will see your own shelves, your own overhead stock and your own price accuracy, not a reference deck.
