Why Most Full Stack Developer Candidates Can't Defend Their Own Numbers

Q: You wrote 'improved load time by 40%'.
How did you measure that?

A: We checked it and it was faster.

Q: Which page? What was the time before?

A: I don't remember exactly. Around 5 seconds
maybe, then 3.

Q: What caused the improvement?

A: We optimised the code.

Result: the line now hurts more than it helps.

Q: You wrote 'cut orders API p95 from about
800 ms to about 300 ms'. How did you measure it?

A: Locally, with a seeded database of my test
orders. I logged response times with a small
Express middleware and ran a short load test.
The slow part was a query without an index on
user_id plus a sort on created_at. EXPLAIN ANALYZE
showed a sequential scan. After adding a composite
index it used an index scan.

Caveat: local numbers, not production.

// Logs how long each request takes (Node.js + Express)
function timing(req, res, next) {
  const start = process.hrtime.bigint();
  res.on('finish', () => {
    const ms = Number(process.hrtime.bigint() - start) / 1e6;
    console.log(`${req.method} ${req.originalUrl} ${res.statusCode} ${ms.toFixed(1)}ms`);
  });
  next();
}
app.use(timing);

// Short load test from the terminal:
// npx autocannon -c 20 -d 15 localhost:4000/api/orders
// autocannon prints latency percentiles, including p50 and p99,
// plus requests per second.

-- PostgreSQL: before adding an index
EXPLAIN ANALYZE
SELECT id, total, created_at
FROM orders
WHERE user_id = 42
ORDER BY created_at DESC
LIMIT 20;
-- Look for: Seq Scan on orders ... and the
-- 'Execution Time' line at the bottom.

CREATE INDEX idx_orders_user_created
  ON orders (user_id, created_at DESC);

-- Run the same EXPLAIN ANALYZE again.
-- Look for: Index Scan using idx_orders_user_created
-- and compare Execution Time with the first run.
-- Use a realistic number of rows, or the planner
-- may choose a seq scan anyway on a tiny table.

// Load the admin page only when it is visited.
// Before: imported at the top, so every user
// downloaded admin code.
import { lazy, Suspense } from 'react';

const AdminPage = lazy(() => import('./pages/AdminPage'));

// In the router, wrap it in Suspense with a fallback
// component, for example a Spinner.

// Measure: run `npm run build` before and after.
// Vite prints each output chunk with its size,
// so you can show the main chunk got smaller and
// a separate AdminPage chunk appeared.

// Send one request after the user pauses typing,
// instead of one request per keystroke.
function debounce(fn, wait) {
  let timer;
  return (...args) => {
    clearTimeout(timer);
    timer = setTimeout(() => fn(...args), wait);
  };
}

const searchProducts = debounce(async (term) => {
  const res = await fetch(`/api/products?q=${encodeURIComponent(term)}`);
  setResults(await res.json());
}, 300);

// Measure: type the same word with the Network tab
// open, before and after. Count the requests.

RESUME LINE: _____________________________

Metric (exact name):    ___________________
Where measured:         local / staging / prod
Tool and command:       ___________________
Data used:              ___________________
Before value:           ___________________
Change I made:          ___________________
After value:            ___________________
Number of runs:         ___________________
Honest caveat:          ___________________

Can I explain this in 60 seconds? Y / N

BEFORE -> AFTER

"Improved performance by 40%"
-> "Added a composite index on the orders query;
    EXPLAIN ANALYZE showed an index scan instead
    of a full table scan"

"Made the app much faster"
-> "Lazy-loaded the admin dashboard so regular
    users no longer download its code"

"Handled heavy traffic"
-> "Load-tested the orders API locally with
    autocannon and fixed the slowest query"

PERF-NOTES.md (keep in your repo)

## Orders API
Date: ____
Command: npx autocannon -c 20 -d 15 .../api/orders
Data: seeded test orders, local Postgres
Before p50 / p99: ____ / ____
Change: index on orders(user_id, created_at)
After p50 / p99:  ____ / ____

## Home page
Tool: Lighthouse, mobile preset, 3 runs
Before LCP: ____
Change: compressed hero image, lazy admin
After LCP:  ____