What a Stock-Counting App Needs: Scanning, Offline and Sync
An inventory stock-counting app for a retail shop needs five core features. These are barcode scanning with a manual fallback, counts that save offline, sync that handles conflicts openly, a variance report that compares counted and expected stock, and a count-session lock that stops the expected figures changing mid-count. Without these, staff end up recounting shelves and nobody fully trusts the final numbers. The rest of this guide explains why each feature matters, the edge cases to plan for, and what happens when two people count the same shelf.
Which features does a stock-counting app actually need?
Plan the features around the count itself, not around a general inventory screen. Here are the essentials, with the edge case that each one has to survive.
| Feature | Why it matters | Edge case to plan for |
|---|---|---|
| Barcode scan | Fast, accurate item identification at the shelf | Damaged, missing or duplicate barcodes |
| Manual fallback | Counting continues when a scan fails | Two products with similar names; staff need search and a confirmation step |
| Offline counts | Back rooms and stores often have weak signal | The app closes before sync; counts must already be saved on the device |
| Sync conflict handling | Several devices may record the same item | Two staff count one shelf and both submit |
| Variance report | Shows where counted and expected stock differ | Items with no count at all, which are not the same as items counted as zero |
| Count-session lock | Keeps expected quantities fixed for the length of the count | Sales continue during the count and move the expected figures |
Each one is covered in more detail below.
How should barcode scanning and manual entry work together?
Scanning is the default way in, but it should never be the only one. Labels peel off, packaging gets crushed and some items arrive with no barcode at all. Give staff a manual route that searches by product name or code and then shows a clear confirmation of the matched item, including its size or variant, before the quantity is entered.
Decide how repeated scans behave. Some shops want every scan to add one unit. Others want one scan followed by a typed quantity, which works better for case packs. Pick one behaviour for each count, or show the current mode clearly on screen, so staff are never unsure whether the last scan was added.
Give an unrecognised barcode its own state with a short explanation and a way forward, such as searching manually or flagging the item for review. Don't make staff guess. Our guide to designing loading, empty and error states covers how to word these moments.
Why must counts work offline?
Stockrooms, cold stores and basements often have poor connectivity. If a count only saves when the network responds, staff lose work or stop to wait. Save every entry on the device as soon as it is made, and sync in the background when a connection is available.
Show the sync status plainly, for example "Saved on this device" or "Synced", so a supervisor can see whether a device still holds unsent counts before the session closes. The wider design choices, including queueing and retry behaviour, are covered in planning offline data sync for a mobile app.
Retries create their own risk. A count sent twice after a timeout must not be added twice. Give each entry a unique identifier so the server can recognise a repeat, as described in preventing duplicate submissions in a mobile app.
How should sync handle two people counting the same shelf?
This is where many counting apps quietly go wrong. If the last upload simply overwrites the earlier one, a full count can disappear without anyone noticing. Before you build, decide which counting model the shop uses:
- Split counts. Each person counts different locations or items, and their entries are added together.
- Blind recounts. Two people count the same location independently, and the app compares the results.
Whichever model you choose, store each person's entries separately with the location, device and time. Then decide whether to add or compare at review, rather than overwriting during sync.
Worked example: two staff, one shelf
The numbers here are illustrative only. Asha and Ravi are both assigned Shelf B3 by mistake. Both are offline in the stockroom.
- Asha scans a pack of batteries and records 12.
- Ravi, a few minutes later, records 11 for the same item on the same shelf.
- Both devices come back online and sync.
If sync used last-write-wins, the system would keep 11 and drop Asha's count with no trace. If it simply added the entries, it would report 23, which is wrong.
A better design notices that two separate count entries exist for the same item and location in the same session. It flags the line as a conflict instead of choosing a winner. The supervisor sees both entries, who made them and when, and asks for a quick recount. The recount finds 12 (Ravi had missed a pack behind the front row), and the supervisor accepts that figure with a short note. Both original entries are kept for the record.
The fix is not complicated. It just requires deciding the rules before development, not after the first disputed count.
What should a variance report show?
The variance report is what the count is for. For each item, show the expected quantity at the moment the session was locked, the counted quantity and the difference. Let reviewers filter for the largest differences, open conflicts and items that were never counted.
Keep "not counted" separate from "counted as zero". A missing count usually means a shelf was skipped. A zero means someone looked and found nothing. Treating them as the same thing produces false shortages.
Let a supervisor request a recount on specific lines from the report, and record who approved the final figure. How the business then accounts for differences is a separate decision for its accountant and is outside the scope of this guide.
Why does a count need a session lock?
A shop rarely stops trading to count. If expected quantities keep changing as sales are recorded, the variance report compares a count with a moving target. A count session fixes the expected quantities when the count starts, for the locations it covers, and records which devices and staff take part.
Decide in advance how to handle sales during the count. Options include counting while closed, locking one section at a time, or recording movements during the session so they can be reconciled at review. Also decide who can open, pause and close a session. Closing should check for devices with unsynced entries first. Our guide to planning mobile app user roles and permissions helps with defining who can do what.
What common mistakes should you avoid?
Relying on scanning alone. Without a manual route, a single missing label stops the count.
Saving only when online. Counts made in weak-signal areas get lost, or staff stop to wait for a connection.
Last-write-wins sync. Overlapping counts overwrite each other silently.
Treating blanks as zero. Skipped shelves show up as false shortages.
No session boundary. Live sales distort the comparison and the variance report can't be trusted.
When does this not apply?
A very small shop with a few dozen items, one person counting and a reliable connection may not need conflict handling or a formal session lock. A shared spreadsheet or a simple list may be enough. Add these features once several people count at the same time, the item range grows or the connection is unreliable. Use the approach in how to scope an app MVP to decide what belongs in the first release.
What questions do founders often ask?
Can a phone camera replace a dedicated barcode scanner?
Often, yes, for small and medium counts. Test it in the shop's actual lighting, with real labels and the phones staff will use. Dedicated scanners can suit very high-volume counts better. Either way, keep the manual fallback, because no scanner reads every damaged or missing label.
Should staff see the expected quantity while counting?
Usually not. Showing the expected figure tempts people to confirm it instead of counting carefully. A blind count, where staff record only what they find, gives a more honest variance report. Supervisors can see expected figures at review, where comparing them with the count is the actual task.
What happens if a device never syncs before the session closes?
Closing the session should warn the supervisor and list the devices with unsent entries. Let the supervisor either wait for those devices to reconnect or close the session and mark the affected locations as not counted. That keeps the gap visible in the report instead of hiding it.
Does every retail app need all five features?
Not on the first day. A single-person count in a well-connected shop may only need scanning, manual entry and a variance report. Add offline saving, conflict handling and session locks once multiple staff, weak signal or trading during counts make those problems real for the business.