Starting point: one code, two entry points
A production plant with gate access control and online safety training wanted each employee to carry one QR badge. The same code arrives in the invitation email, is scanned at the gate, and is meant to open the training on a tablet in the room by the entrance. The safety manager expected the system to know, after a scan, who stands in front of the tablet.
In practice employees reported the opposite. Training was marked as completed, yet the gate still showed no valid training. Security sent the person back to the room, the safety manager searched the data for the cause, and the employee lost time at the entrance.
Why the training landed on another account
The diagnosis pointed to two mechanisms. First, the phone scanning the printed badge opened an address with a token, but the token was dropped on the way to the training reader. The tablet did not know whose code had been scanned, so it showed a form with company and login fields.
The second mechanism: the form accepted any login, including one with a typo or someone else's login. Earlier, the training login created accounts silently, so a typo produced a new, phantom employee. Training completed on such an account did not exist for the gate, which checked the account linked to the badge.
- the token from the QR code did not reach the training reader,
- the company plus login form allowed selecting an account other than the one on the badge,
- the login created an account instead of first searching for an existing one.
Steps to handle one case of "training passed, but the gate does not let me in"
How it was closed: changes in logic, not in people
The first change concerned passing the token. The phone scanning the badge opens an address with the token, and the system forwards it to the training reader without losing it. The tablet therefore receives the identity from the same code the gate scans.
The second change: when training is entered from a badge, the company plus login form is not shown at all, even if the code is not recognised. The ability to type someone else's login was the very flaw, so it was removed instead of being limited with a warning.
The third change: the training login no longer creates accounts silently. The system first searches for the existing account and logs in by its identifier. A typo ends with a message, not a new account.
A scan does not log in: the person confirms identity
Scanning the QR code on the training tablet logs nothing in. The tablet shows the first name, last name and company of the badge owner, and training starts only after the person confirms with the question "Is this you?". This protects against a situation in which an employee scans someone else's, a found, or a borrowed badge.
When a badge is used for identification, the system never returns the scanned code or token, only the identity to confirm. An unknown token, a revoked badge and a made-up string of characters all receive the identical 404 response, so nobody can tell from the response which codes exist. The token itself has 256 bits from a random generator, and the QR code encodes the instance address with the token, so guessing codes is not a practical attack.
One code recognition rule for the gate and the tablet
One server-side rule, shared by the gate and the training tablet, determines what the scanned string is. The browser holds no copy of it, so the format cannot drift after a change. As a result, a fix to the code format applies to both places at once and does not require updating the tablets.
The security panel accepts a scan of a QR badge or an RFID card and returns a verdict made up of valid trainings, qualifications for hazardous work, an active permit and an active account. The scan itself records the entry. After the changes, the gate verdict is based on the account that receives the training from the tablet.
Result and conclusions for the safety manager
After the fixes, the question "which account was the training recorded on" no longer has two answers. Reports of the type "training passed, but the gate does not let me in" now concern only real gaps: training past its due date or an unfinished test. The example calculation below shows how many steps the team saves for one person.
General conclusion: when one identifier is to serve several touchpoints, first establish which of them only read identity and which change it. Connect the gate and the tablet with a shared server-side rule, and leave the person one clear decision: "it is me".
Example calculation: time for one case
| Parameter | Value |
|---|---|
| Steps before the change (referral, account search, manual correction, repeat scan) | 6 steps |
| Steps after the change (notice of the gap, referral to training) | 2 steps |
| Time per step | 3 min |
Result: 18 minutes before the change and 6 minutes after the change for one case, that is 12 minutes less.
Note: Example calculation based on the assumptions above; actual time depends on gate organisation and the number of cases.
Video
Frequently asked questions
Does a scan of the badge on the tablet log into training automatically?
No. The tablet shows the first name, last name and company of the badge owner, and training starts after confirmation "Is this you?". This protects against the use of someone else's badge.
What happens when a revoked badge is scanned?
The system responds the same way as for an unknown code: 404. The person cannot tell whether the code ever existed, and a revoked badge does not pass as valid.
Can a typo in the login create a new account?
No. The training login first searches for the existing account and logs in by its identifier. When entering from a badge, no login is typed at all.
Does a change to the code format require changes on the tablets?
No. The code recognition rule runs on the server and is shared by the gate and the tablet, so the browser does not keep a copy of it.
Want to close the gap with one QR badge for the gate and training?
We will show how the same QR code serves entry and training at your plant. Book a conversation or a demo.
Let's talk