Decision for the CEO · in-app search
Buy hosted search. Don’t build it.
Better search for customers in up to 6 weeks, not a quarter.
Approve buying the hosted search service with its EU region, so procurement can start now. This commits about $57.6k a year and one engineer for 3 weeks. It does not rule out building our own later.
Search drives nearly a quarter of detractor complaints
Search is our second most common NPS complaint.
- 1. Speed
- Detractors are complaining now, so the time to a fix comes first.
- 2. Engineer time
- Every engineer-week spent here is taken from other work.
- 3. Cost at scale
- Records grew 2.4x last year.
- 4. Ranking control
- It matters, but less than shipping a fix.
Buying ships sooner with a thirteenth of the engineering
Build: 3 engineers for a quarter (about 13 weeks). Buy: 3 weeks of procurement, then 1 engineer for 3 weeks; less if the two overlap.
Building wins on ranking control; buying wins elsewhere
| Criterion | Buy hosted service | Build our own |
|---|---|---|
| Speed | Up to 6 weeks | About 13 weeks |
| Engineer time | 3 engineer-weeks | 39 engineer-weeks, then half an engineer |
| Cost at scale | $57.6k a year; about $108k at triple the records | Half an engineer, ongoing |
| EU and security | EU region; SOC 2 Type II | We host in the EU ourselves |
| Ranking control | What the vendor allows | Full control |
Prices include the 20% EU charge. If records keep growing 2.4x a year, they triple in about 15 months.
Approve buying hosted search so procurement starts now
- This commits
- About $57.6k a year including the EU region, plus 3 weeks of one engineer’s time.
- What we give up
- Our platform lead’s strongest point: full control over ranking.
- If we’re wrong
- We can still build later; we lose the 3 integration weeks and the contract term.
- What would change this
- Evidence that the complaints are about ranking order and the vendor’s settings can’t fix it.