No per-person surface
There is no roster, no per-person report, no "online now," and no per-unit-time figure anywhere in this app — not to you, not to a customer, not in an export. What the operator does not get is what makes the record worth having.
One thesis: this app has no per-person surface, and that absence is the product. You can ask who touched this pallet and get an attested answer. You cannot ask what has this person done, and neither can anyone else with access to your deployment. Object → performer is the product; performer → objects is refused. Every other refusal in this post follows from that one.
What you were expecting
A frontline app usually arrives with a people view. A roster of who is on shift. A per-person count of completed tasks. A picks-per-hour figure, maybe with a target and a color. An export to spreadsheet with one row per worker.
You will not find any of it here, and this is not a gap in a first version. Here is the full list of what is absent, so nothing has to be inferred:
- No roster. No list of people anywhere in the app, for anyone. There is no assignee column on a worklist, because every row is already the reader's own.
- No per-person report. No query that returns two performers side by side, in any face, in any export.
- No per-unit-time figure. No duration, no picks per hour, no time in queue, no shift progress bar, no "ahead of / behind," no streak. Allowed: a count, a step indicator within one task, and a timestamp.
- No presence. No "online now," no connected count. A transport that can enumerate who is on shift is a roster arriving as an implementation detail, so the transport does not expose one.
- No supervisor face. A supervisor who opens a deployment is a performer in it or is nothing in it.
- No target, no scoreboard, no badge, no level. Turning a per-unit-time figure into a game is still a per-unit-time figure.
Why, in one sentence
Because the thing you are paying for is an attested who on every event, and the attested who is only there as long as the people supplying it have no reason to make it flattering.
The moment an app will answer what has this person done, every person on it has a reason to make the answer look good, and the cheapest way to do that is to stop scanning honestly — to scan the easy case twice, to skip the one that will be slow, to close on the tap and not the pallet. A per-person report poisons the ledger it is drawn from. Your recall record on a lot of lettuce is made of that ledger. You cannot have both.
What you do get, fully
Who touched this pallet. Ask it of any object — a case, a pallet, a lot under recall — and the answer is every attested act against it: the receiver at dock 12 at 06:14, the inspector at 06:40, the approving shift lead at 06:52. Names, places, times, event hashes. This is the question an auditor, a customer and a chargeback panel ask, and it is answered without reservation.
Who holds this task. A task's own record shows its assignments and its current holder. That is the object → performer direction, and it is permitted.
A count. How many tasks are pending at this station. How many were closed against this PO. A number about work, not about a person.
Where the line is enforced
Not in the screen layout, where a column could be added next quarter. The work door behind this app, worklists.dev, refuses the performer → objects direction where a query is constructed — in the row grain, in the group-by, in the parameters by name and by value, and in the keys a face would render. A request for your deployment's data that groups by person does not return a permission error. It returns a typed refusal, because the query cannot be built.
And a worklist query has no field for an assignee at all. "Me" comes from the login. There is no way to ask for someone else's list by writing a clever request.
Identity is regime-scoped
Where a regulation requires the performer be identified — food-safety monitoring under HACCP, forklift operation under OSHA, and the like — who-completeness on those events is a compliance metric, and it is legitimate because the regulator set it. Where nobody requires it, it is not sold. A scope with no cited authority does not get the metric. Your deployment's regime is compiled in, and the screen tells each person which one they are under.
That is what keeps a frontline app from becoming the whip, and it is what lets the people on your co-packer's floor scan into it as if it were a record rather than a trap.
What to tell the person who asks for the report
That the report and the evidence cannot both exist, and the evidence is what the customer's auditor will accept. Get started with an evaluation deployment, try to find a list of people in it, and then ask it who touched a specific pallet. One of those comes back.
The record of work is the evidence it happened →
One motion closes a task on this app — scan, attest, emit, complete — and the record it leaves is not a log entry about the evidence. It is the evidence. Who, where, when, inside the window, in one object.
