01
Not a single photograph
- Problem
- Sales has no subject. Any image would have been decoration.
- Decision
- The effect comes from numbers counting up and from the search underneath them.
- Why
- Somebody who has used the tool no longer needs the picture.
← All work
Fluxrate — Sales & lead generation
A lead agency sells something there is no picture of. So the tool itself had to go on the page.
Starting point
Every sales agency page claims the same thing: more enquiries, better meetings, predictable growth. None of it is checkable, which is exactly why nobody believes it. The only way out was to let the visitor try it.
The three questions, before the first line of code
01
What does the business do?
Lead generation for local businesses.
02
Where?
The Ruhr region, working remotely.
03
What should the visitor do?
Run the search once. Somebody who has seen results asks a different question.
Decisions
02
The real search, not a rebuilt one
- Problem
- A simulated search with canned results is exposed in ten seconds — people type in their own town.
- Decision
- The same Google Places query that runs in the internal lead scout, through a public route of its own.
- Why
- The risk is abuse, not disbelief. Abuse can be capped; lost trust cannot.
03
Rate limiting without an admin key
- Problem
- A public route that costs money per call, and no service-role key, which has no business in a browser.
- Decision
- Two security-definer functions in Postgres. The IP is never stored, only a hash made with a secret key.
- Why
- The same construction already ran for the public site analysis. A second, separate mechanism would have been a second thing to get wrong.
What actually runs
/api/leadfinder-public, three searches per IP per day, fifty overall. The route never writes to the real leads table — the study and the internal tool share the query, not the data.
What I would do differently
I would also sort the results by who has no website at all. That line is what decides things in the internal tool, and it is missing from the public version.
Next case study
Glanzform →