The whole
argument,
in order

Substandard and falsified medicines are a distribution failure, not a manufacturing one. Product leaves legitimate factories correctly; it is imitated, diverted, mishandled or degraded on the way to a person. Every existing control sits upstream of the counter, and the counter is where the chain of custody ends.

Consequence: the harm is invisible until it is a death, a treatment failure, or a resistance pattern. Nobody is watching the last metre.

The patient is an unused sensor. They are present at every dispensing event, they observe the pack directly, and they experience the outcome. What they lack is not motivation but an instrument and a destination for the signal.

Give them something independently useful, guidance in their language, offline, free, and the reporting channel comes for free with adoption. Build the reporting channel first and nobody shows up.

A progressive web app rather than a native download: no store friction, no install cost on a constrained device, and it survives on cheap Android. Four capabilities, three of which are live today.

• Reference, safety and adherence guidance in five languages, no dose figures. Live.
• Offline, full service-worker caching. Live.
• Reporting, device-local suspect-pack reports with a reference code. Live.
• Adherence, local dose reminders. Live.
• Pack verification, computer vision against manufacturer data. Not built, and not simulated.

Now, usefulness without partners

Reference, offline, reporting and reminders, shipped and honest. Clinical and translation review of every entry.

Next, the shared map

A moderated backend with rate limiting turns device-local reports into geographic clusters. First pharmacy pilots in one state.

Then, the upstream deal

Cluster data is the argument that brings a manufacturer to the table. Their serialisation data is what makes verification real.

Later, the regulator feed

Formalised signal sharing, and cross-border expansion where the same failure repeats.

The honest risks, stated as conditions rather than hidden as assumptions.

• People will use a reference app often enough for reporting to reach useful density.
• A moderated report stream resists defamation and competitor abuse well enough to be credible.
• At least one manufacturer will share serialisation data before verification can exist.
• Clinical reviewers can be secured at a cost that scales with the reference set.
• Regulators will accept crowd-sourced signal as an input rather than a nuisance.

Any of these failing changes the plan. None of them are assumed silently.