BACKEND TEARDOWN
TWELVE QUESTIONS ABOUT YOUR BACKEND. YOU GET A SCORED REPORT — THE FAILURE MODES YOUR ANSWERS POINT AT, WHY EACH ONE COSTS YOU, AND WHAT I WOULD FIX FIRST. FREE, NO CALL REQUIRED.
What you get
- A score out of 100 across five areas: data layer, caching, reliability, security and delivery.
- A prioritised list of findings — each one with why it matters and the concrete fix.
- An estimate of where this design stops holding up, in requests per day.
- The full report emailed to you, if you want it. The score itself needs no email.
Roughly 90 seconds. Every question has a “not sure” option where it makes sense — an honest unknown scores better than a guess.
What the teardown looks at
A backend rarely fails everywhere at once. It fails at its weakest layer, under load, usually at the least convenient hour. The teardown scores five layers separately so you can see which one is yours.
Data Layer
Schema, queries and the indexes holding them up.
Performance & Caching
What you recompute that you could have remembered.
Reliability & Ops
Background work, and whether you find out before your users do.
Security & Access
Who can call what, and how hard it is to abuse.
Delivery Confidence
How safely a change reaches production.
The questions
Nothing is hidden — these are the twelve questions, in order. Every scored answer maps to a known failure mode, and the report tells you which ones your answers triggered.
- 01What is the backend built on?
- 02Roughly how much traffic does it handle?
- 03What is the primary database?
- 04How does your code talk to the database?
- 05Do you know which queries are your slowest?
- 06What caching is in place?
- 07Do you know your p95 API response time?
- 08How is background work handled — emails, exports, webhooks?
- 09How do you find out when production breaks?
- 10How is authentication handled?
- 11What stops your API from being abused?
- 12What happens between a merge and production?
- 13What brought you here?
How the score works
Each scored question contributes to one of the five categories. Categories are then weighted rather than averaged flat — a reliability or security gap costs more than a manual deploy, so it moves the headline number more.
Answers of “not sure” score below the middle rather than at zero. Not knowing is a genuine risk, because you cannot fix what you cannot see, but it is not the same as knowing the answer is bad.
A critical finding caps the overall score regardless of how tidy the rest is. A system that cannot detect its own failures does not get to be called healthy.
Questions about the teardown
What is a backend architecture teardown?
A backend architecture teardown is a structured review of how a backend is built — its data layer, caching, background work, security boundary and deploy process — that identifies which part will fail first as traffic grows. This one is questionnaire-based: you answer twelve questions and get a scored report with prioritised findings.
Is the teardown really free?
Yes. The score and findings appear on the page as soon as you finish the questions, with no email required. You can optionally have the full written report emailed to you.
How long does it take?
About 90 seconds. Twelve single-click questions, each with a 'not sure' option where not knowing is a legitimate answer.
How accurate is the score?
It is a directional read, not a code audit. The scoring engine flags the failure modes your answers point at — it has not seen your repository, so treat it as a prioritised list of things to check rather than a verified diagnosis.
What does the teardown check?
Five areas: the data layer (database choice, query patterns, slow-query visibility), performance and caching, reliability and operations (background jobs, monitoring), security and access (authentication, rate limiting, input validation), and delivery confidence (tests and deploys).
Do I have to book a call afterwards?
No. The report is yours either way. If you want to go through the findings against your actual codebase, there is a free 30-minute call available, but nothing about the report depends on taking it.
Prefer to just talk it through? The contact page reaches me directly, and the case studies show what these fixes look like in production.