The poll workers nobody can verify
In the weeks before its general elections of 24 June 2023, the Electoral Commission for Sierra Leone (ECSL) recruited and trained roughly 91,300 people. They were sent to 11,832 polling stations spread across 3,630 centres, most of them a long way from the nearest district office. Hardly any had worked for the commission before, and most would never work for it again.
That is an ordinary number. Every election management body (EMB) in the world assembles a temporary institution of this kind, runs a country through it in a single day, and dissolves it within a fortnight. What very few of them can do is answer a plain question about it: when that worker sits down at the station on the morning of the poll, how do you know it is the person you hired?
In most commissions the honest answer is that you trust the list. A presiding officer checks a name against a printed register; a supervisor recognises a face. Neither survives a substitution, a mislaid register, or a station the supervisor never reaches. And the question does not go away when the polls close, because the identity nobody confirmed at the door has to be confirmed all over again before anyone is paid.
The bill arrives late
Sierra Leone's payment round is instructive. Funds for temporary polling staff moved through district bank accounts financed by the UN Development Programme, and the commission was still publicly urging staff to come forward and claim their money in October, more than three months after polling day. Some of that is the ordinary friction of disbursing cash through a banking network that thins out sharply outside Freetown. But a share of it is verification: before a commission can pay 91,300 people it barely knows, it has to satisfy itself that each of them is who they say they are, worked where the roster says, and has not been paid already.
The commissions willing to say all this out loud tend to be the better-run ones. Naming an operational constraint precisely is the first condition of fixing it, and the constraint here is unusually precise. It is not fraud, or at least not mainly. It is that the workforce is large, temporary, scattered and assembled at speed, and the tools for confirming who is in it were designed for offices with desks and broadband.
Why the obvious fix is not one
The instinctive response is that there must be an app for this. There are many. Almost none of them work where the problem lives, for three unglamorous reasons.
The first is handsets. The temporary staff an EMB most wants tend to be older, rural and trusted in their own communities, which is close to a description of the people least likely to own a smartphone. A system that assumes one in every pocket quietly disqualifies the best recruits.
The second is signal. Polling stations are, by design, in the places national networks reach last. Sierra Leone has spent some $400m on telecommunications infrastructure and extended its fibre backbone to 14 of 16 districts, and internet penetration still stood at 20.8% of the population at the end of 2025. Roughly four people in five were offline. The gap between network coverage on a map and usable connectivity at a rural polling station is where most electoral technology goes to die, and it is a usage gap as much as a coverage one: affordability and devices, not only masts.
The third is time. Temporary staff are hired weeks before a poll, not months. Any scheme requiring a long enrolment exercise, a stable national database or a serious retraining programme will not be ready, and the budget line to build one does not exist.
The boarding-pass principle
The underlying problem, proving that someone is who they claim to be without asking a server at the moment of the check, was solved some time ago. It is why a boarding pass scans in airplane mode. The proof is carried inside the code, cryptographically signed by the airline when it was issued, and the gate reader confirms the signature against a key it already holds. Nothing is fetched. Nothing needs a network.
The same architecture transfers to a poll worker's identity without much imagination. A credential is issued once, at recruitment, when the commission does have connectivity. It travels as a QR code, on a phone or printed on card, carrying the holder's photograph, station assignment and role, signed by the commission. A supervisor with a single device verifies the signature offline, at the door, and the device syncs its log back to the district office whenever it next finds a connection. Attendance and hours are then attached to a confirmed identity rather than a handwritten name.
None of this is speculative. Humanitarian registration systems have run identity verification in disconnected settings for years, and the international standard for mobile driving licences, ISO 18013-5, was written specifically to allow a credential to be checked with both devices offline. The card in a European wallet app works on the same principle. What has not happened is anyone building it for electoral workforces, which are an awkward market: enormous for six weeks, non-existent for four years.
The printed variant matters more than it sounds. It converts the requirement from every worker owning a smartphone to one working device per cluster of stations, which is an entirely different procurement conversation.
What it would not fix
Commissions considering this should be clear about its limits. A verifiable credential proves that the bearer holds something the commission issued. It does not prove the commission issued it to the right person; that depends entirely on the quality of the identity check at recruitment, which is the step most likely to be rushed. Nor does it prevent a credential being handed to somebody else, unless it is bound to the holder by a photograph, a PIN or a biometric, each of which adds cost and training.
There is a records question too. A system of this kind creates an authoritative dataset of who worked at which station and when, which did not previously exist in retrievable form. In jurisdictions with strong public records law, that dataset is potentially disclosable, and officials are entitled to ask who holds it and under what terms before they agree to be enrolled in it. Commissions should settle that before procurement, not after.
For EMBs weighing it up
Start with the payment problem, not the door. Payroll verification has a hard number attached to it and a finance director who cares. Attendance checking is easier to justify once the payment case is made.
Test the connectivity assumption on the ground. Coverage maps are drawn by network operators. Send someone to three of your most remote stations with the device you intend to use, in the week before a poll, and see what happens.
Insist on offline verification, not offline caching. These are different claims. A system that caches a list and synchronises later still depends on that list having reached the device. A signed credential does not.
Settle the data questions first. Who holds the credential record, for how long, and who can compel its disclosure. An official who cannot get a clear answer will decline to participate, and will be right to.
The Network is gathering members' experience of workforce verification ahead of the next round of national polls. Commissions that have tried something here, successfully or otherwise, are invited to get in touch; accounts of what did not work are as useful as accounts of what did.