I run a small data service, and before I started charging for it, I needed to prove the pipeline worked. The honest version of "how I built a competitor price monitor" is a lot less impressive than the tweets make it look. No ML. No orchestrator. Just three dumb pieces that hold together:
- A scheduled fetch that grabs a handful of public pricing pages.
- A regex pass that pulls out the dollar amounts and tags them with the date they were seen.
- A model that turns the raw diffs into a short "what actually changed" note - not a prediction, just a dated record.
The part that took the longest wasn't the code. It was deciding what "changed" means. A page re-renders and the same $19 shows up again - that's not a change. Price went $19 ? $25 on Tuesday - that is. Filtering noise, keeping a dated audit trail, and not claiming more than the data shows: that's 80% of the job.
I put together a sample of what a real weekly report looks like (scraped three real public pricing pages today, dated). If you've ever wanted eyes on a competitor's pricing but didn't want to build this yourself, that's the kind of thing I do now - drop me a note and I'll run a free 7-day snapshot on any 2-3 pages you care about, no setup on your end.
This is the honest middle: the tool isn't the hard part, the discipline around "what actually changed" is.
Note: drafted with AI assistance, reviewed by a human. All data in the sample is real, scraped from public pages today.
Top comments (0)