What Happens When You Check an A/B Test Early
Twenty thousand simulated experiments with no real difference in them. Read once at the end, 4.9% came back significant,...
Guides & How-tos · Worked examples · Real numbers
Step-by-step how-tos, head-to-head comparisons and honest roundups — every guide worked through with current figures, then backed by a RECATOOLS calculator or AI-directory entry so you can act on it.
Twenty thousand simulated experiments with no real difference in them. Read once at the end, 4.9% came back significant,...
Metadata queries are expensive because the server assembles the answer by walking itself. Measured on production with th...
Knowing a table is scanned tells you little; knowing which statement scans it tells you everything. Reading MySQL's stat...
Indexes accumulate because nothing ever removes them, and removing one means proving it is unused. MySQL will tell you:...
MySQL 8 ships 89 collations for utf8mb4 and exactly one is for a Southeast Asian language. Five of them sorted the same...
Truncation looks like the simplest thing a program does, until the same twenty-character limit hands an English reader n...
Fifty-nine thousand runs, forty-four commands, zero failures — because for background tasks the failure path ran in a pr...
A naive date filter and a timezone-aware one disagreed on twelve of twenty-one days, once reporting six items on a day t...
Five clients, one job, one throttle. Hammering issued 553 requests to fetch 40 things and finished behind the client tha...
With the database stopped the homepage still returned 200, at a fifth of its normal text with none of its articles. Find...
Twenty questions with known answers, run over 205 real documents. Prefixing titles to chunks raised hit@1 from 14 to 17...
We tried to measure our own monthly egress with root on the machine. Three sources gave three answers, and extrapolating...