When they install a sticker, my poseurs — the installers in the field — need to identify the exact shop they're standing in front of, fast. Typing the name by hand on a sidewalk is the bottleneck. What's the cleanest way to make finding the right shop almost instant?

Before designing anything, one question decides the rest: what does the phone give us at that moment, and what are you willing to store? Concretely: do we have GPS and the camera, and can we keep shop data in your own database?

I have two things to work with:

  • The poseur's GPS position.
  • The camera.

That's it. The goal is simply to make the process easier.

Then GPS is the key. Three ways to use it, from most direct to most manual:

  • Live Google autocomplete, guided by GPS. Inject the last known position into Google Place Autocomplete (or the JS widget) so nearby shops appear first, weighted toward your target product types. Easy to build, but you pay Google for every lookup and it stops working when the signal is weak.
  • Capture now, match later. Save the raw GPS position at installation time and match it to shops afterwards, in batch, using other APIs. Cheap in the street, but the poseur confirms nothing on the spot.
  • Photo only, manual. The poseur photographs the storefront and you match it by hand later. Works even for shops Google doesn't list, but slow for everything else.

My default would be the live autocomplete: one tap, right there. But it depends on Google for every single lookup, and that's the part worth questioning.

Here's what I'd rather do. The poseurs know their tournées in advance: each morning or afternoon, they cover a fixed set of streets. So let's process those streets ahead of time:

  • Pre-load the shops Google knows on those streets into my own database, with their type.
  • At installation time, no live Google call: I query my internal database by GPS, with product-type filter buttons.
  • If nothing shows up within a 50-60m radius, fall back to a Google search, then to manual entry with a photo.

That removes the per-lookup cost and works even with a bad signal.

That's the right choice, and it turns the cost problem around: you pay Google once, up front, instead of on every lookup. The pipeline, in priority order:

  1. Process the streets ahead of time. Before each tournée, pull the shops Google lists on those streets and store them locally, with their type.
  2. Rank by GPS. At installation time, read the phone's position and show the nearest shops, filtered by the product-type buttons. No network call needed.
  3. Fallbacks. A Google search if the shop isn't within the radius, then manual entry with the GPS position attached.

And you get fraud control for free: every installation records its GPS position, and the photo proves the poseur was really in front of the shop, so an installation logged from the other side of town gets flagged.

Perfect. Let's build it.

Send an email to Eliott

Eliott Baylot's curriculum vitae

Claude is an AI and can make does make mistakes. Please triple-check responses.