General

A plain-English breakdown of statistical significance, p-values, and sample size, explained with a coin flip instead of a formula, for anyone testing changes in a small business.

A plain-English guide to running an honest A/B test on small traffic, and the specific ways solo founders trick themselves into calling a coin flip a win.

Fourteen months on I can still open the file, just not read it: a column called status holding three values I can no longer name. Six lines of plain text beside a dataset is what keeps it usable.

A plain-English guide to building a cohort retention table by hand, why blended churn numbers hide real trends, and how to read the shape of a cohort curve.

A plain-English walkthrough of customer lifetime value: the formula, a worked example, margin adjustments, the CAC ratio, payback period, and the common ways the number gets misused.

A chart can change what it appears to say without a single number changing underneath it. The axis, the smoothing, the colour and the missing denominator are all decisions somebody made, and for your own dashboard that somebody is you.

A plain-English guide to monthly recurring revenue and annual recurring revenue: what each one measures, why ARR isn’t just MRR times twelve, and when to trust which number.

There is no row count that makes a number trustworthy. Enough depends on the size of the difference you are chasing, and a large sample of the wrong people is worse than a small sample of the right ones.

A parsing bug hid in my price scraper for 19 days. The fix took one afternoon, and it only took one afternoon because the raw pages were still sitting on disk untouched.

Lagging indicators report a result you can no longer change. Leading indicators move first. Here is how to pair a couple of each without getting fooled.