Custom development
Every order had an address and none had coordinates. Once they were mapped, one account turned out to be 20% of the book and the default view was hiding another 35%.
01Key numbers
02The situation
The firm had 254 orders carrying a shipping address. None of them had coordinates, so nothing was on a map and no territory review was possible without asking someone to build a report.
Customer concentration risk was therefore invisible. Not disputed, not estimated, simply not visible to anyone.
Recognise this in your own org?
03What we built
A Lightning Web Component over a map, with geocoding and model work moved off the request path and onto a schedule.
New orders geocode on save through a trigger, the back catalogue geocodes in batch, and a nightly queueable chain fills the gaps. A wrong location is worse than an empty one, so there is no any-province-to-its-country guess.
The browser Maps key sits in a custom setting, the server geocoding key in config, and the model key in a named credential password. In an unmanaged org, protected visibility on a custom setting restricts only managed-package subscribers, so it is not access control there.
Apex allows 120 seconds of cumulative callout time per transaction. The chain takes four orders per execution with 25, 45 and 60 second timeouts and a 90-second wall-clock stop. It is built never to reach the limit rather than to recover from hitting it.
The first full run analysed 2 of 256 orders and stamped 250 with a rate-limit error, because the free tier caps the model at 20 requests per project per day. Generation moved to a hosted worker. Prompts, inputs and the cache rule stayed in Apex so the two paths cannot drift.
Percentages are of order count, not value. The book carries two currencies as entered and is not converted.
We will demonstrate any of these live, on the real org,.
89 of 254 orders sat in statuses that the default filter excluded. Nobody had chosen to hide them. Finding that was worth more than the map itself, because the map was only the instrument that made it visible.
04Outcome
The map shipped with geocoding complete, insight generation running nightly, and an alert that fires when it stops.
An adversarial review produced 30 finding reports, de-duplicated to 16 distinct defects. Fifteen were fixed, one partially, and 14 further findings were investigated and rejected.
Keep reading
The client told us the tax was calculating by product instead of by province. We tested that against all 2,335 order line items and it was not true. Six other defects were, and we priced each one.
Field crews were opening five or six systems before starting work, and a logging flow was telling them it had saved data when it had saved nothing.
Tell us what is broken. You get a written read on it within one business day, before any money changes hands.