Shelf recognition inside your Salesforce Consumer Goods Cloud
A visit runs in Consumer Goods Cloud. When it calls for a shelf check, image recognition opens with the store and the category already set, measures the shelf, and writes the result back to the Salesforce record through the API.
Objective shelf data on the account record, from the visits your reps already make. Running in production today at a multi-country food manufacturer.
Between the tick box and the shelf
What your CRM knows about the shelf is somebody’s opinion
A CRM holds the account, the visit, the tasks and the order. When the visit includes a shelf check, what it can hold is a set of answers: yes, the planogram was followed; roughly this many facings; the promotion was up.
Those answers are a judgment made by a person standing in front of a shelf with fifteen minutes and a phone. Two reps in two stores will not judge the same shelf the same way, and neither of them is being careless about it.
Photographs do not solve it on their own. A visit that ends with six pictures attached to a record leaves the work undone, because somebody still has to look at them, and nobody has time to look at all of them.
So the shelf reaches head office as a number that cannot be compared across stores, and it reaches it after the rep has already driven away. Whatever was wrong on that shelf stays wrong until the next visit.
How it works
One loop: Salesforce, shelf, Salesforce
A visit with a shelf check runs the same way every time, in these four steps.
In Salesforce
The visit starts where it always starts
The rep opens the account in Consumer Goods Cloud and runs the visit: tasks, order, everything the CRM already does well.
Recognition opens
The shelf check starts with context
Salesforce hands over the context: which store is being visited, which category is being checked and what should be on that shelf. Image recognition opens inside that context, so the rep picks nothing from a list and no audit lands on the wrong store.
The shelf gets measured
Photos in, measurement out
The rep photographs the shelf. The engine reads it, the AR layer points at what is missing, and the rep can fix it while still standing there. This part works with no connection.
Back in Salesforce
The result lands on the record
When the audit closes, the result returns to the Salesforce record through the API. The rep carries on in the CRM they started in.
The audit itself opens in our application, and the flow is designed that way to give the rep the best of both tools. The camera, the AR guidance and the offline mode run where they work properly, the CRM keeps the visit and the order, and the finished measurement lands back on the Salesforce record.
Choosing the numbers
We measure the whole shelf. You choose what comes back.
The engine measures more of the shelf than most teams need in one year. You choose the numbers that move your targets, and those are the ones that travel back to the Salesforce record.
MissingThe products that should be there
Obligatory listings and the line you launched last month, checked product by product against what that store is supposed to carry. A gap comes back to the record with the product name and the store attached.
The price on the label
Printed prices are read off the photograph and held against the price you recommended for that product. A store sitting above or below it arrives as a number you can take to a conversation.
Every position against the planogram
The arrangement on the shelf held against the planogram in force, position by position. A competing product standing in space you paid for shows up as a flagged position on the record.
The same audit also returns facings, share of shelf, category length and whether the promotion is standing. Which of them land on the record is your call, and it changes when your targets change.
Here is one client’s set, as it arrives on the account record.
Illustrative example, not client data.
Per product, as written to the visit record
| Product | Listed | Found | Facings | Share of shelf | Status |
|---|---|---|---|---|---|
| Product A, 1 l | yes | yes | 14 | on shelf | |
| Product B, 0.5 l | yes | yes | 11 | on shelf | |
| Product C, multipack | yes | yes | 9 | below agreed space | |
| Product D, new line | yes | no | 0 | missing | |
| Product E, seasonal | yes | yes | 7 | on shelf |
Comparing engines
Three questions worth asking any recognition vendor
When does the result arrive?
Ours is finished while the rep is still in the store, so the shelf can be corrected on the spot. An engine that returns the reading overnight turns every fix into a second visit.
What happens with no signal?
Back rooms and chilled aisles lose connection. Our audit runs anyway and syncs when the phone finds a network again, so a dead spot costs you nothing.
Cola regular 0.5 L bottle
Does it tell the rep what to do?
The AR layer marks the gaps on the camera view while the rep is still in front of the shelf. A report that lands in head office leaves that work to somebody else.
The division of labour stays simple. Salesforce keeps the account, the visit and the order. We measure the shelf and hand the numbers back to it.
How good the reading is
A recognition engine with numbers behind it
The integration is the easy part to talk about. What decides whether the numbers are worth putting in front of a retailer is the engine behind them, and ours has been reading shelves in FMCG for years.
Getting connected
What it takes to connect it
Your side
You have Salesforce
Consumer Goods Cloud, in whatever shape you run it today.
Our side
We configure the connection
Every Salesforce estate is set up a little differently, so the connection is configured to match yours. Configuration work, not a development project.
Together
You choose the scope
Which categories, which indicators, which stores go first, and what the result should look like when it lands on the record.
Not sure which numbers you need yet? The Retail Image Recognition page covers what the engine reads. If you already know what you want written back, tell us the category and we will scope it against your estate.
Questions
Image recognition inside a Salesforce estate
Does Retail Image Recognition integrate with Salesforce Consumer Goods Cloud?
Yes. Image recognition opens from a visit in Consumer Goods Cloud, the shelf audit runs, and the result returns to the Salesforce record through the API. The integration has been delivered and runs in production for a multi-country food manufacturer.
Does the rep leave Salesforce?
For the audit itself, yes. Recognition opens as its own step, because that is where the camera, the AR guidance and the offline mode live. The rep then continues the visit in Salesforce, with the measurement already on the record.
What data comes back into Salesforce?
Whatever you choose. Availability, facings, share of shelf, planogram match and promotion compliance are the common ones. The set is agreed when the connection is configured.
How accurate is the reading?
Up to 98% recognition accuracy, against roughly 80% from manual self-reporting. What a given category reaches depends on the photograph and on how well the product catalogue is prepared, which is part of the set-up work.
What syncs from Salesforce into image recognition?
Stores and their attributes, users and the org hierarchy, products and catalogues, the assortment each store should carry, and the rep to store assignment. Salesforce stays the system of record.
What do we need in place before connecting it?
Salesforce on your side and a configured connection on ours. Estates differ, so the connection is set up to match yours. It does not need a development project.
Do we have to replace our CRM?
No. Salesforce stays your system of record; we add the shelf measurement and hand it back.
See it on your own Salesforce
Bring one category and a handful of stores. We will run the loop end to end: the visit in Consumer Goods Cloud, the audit on a real shelf, and the result written back to the record.
Book a demo