Numbers that describe a business, and numbers that flatter it
Every metric is a compression of reality, and compression discards something. The skill is knowing what each number dropped on its way to the slide.
Growth rate without retention describes a bucket filling and leaking at once. Revenue without gross margin describes activity, not a business. Averages conceal distributions: if a handful of accounts carry most of the revenue, the average customer does not exist and planning around it produces decisions that fit nobody.
Pairs that have to be read together
| Metric | Meaningless without | What gets hidden otherwise |
|---|---|---|
| Growth rate | Retention | Churn replaced by new sales |
| Revenue | Gross margin | Reselling someone else's cost base |
| Acquisition cost | Payback period | Cheap customers who take three years to repay |
| Averages | Distribution | Concentration risk in a few accounts |
| Conversion rate | Sample size | Noise read as a trend |
Sample size, and the temptation to read noise
Small numbers move a lot on their own. A test that converts 4 of 50 against 7 of 50 looks like a large improvement and is well within what chance produces. Acting on it is not merely harmless optimism — it spends the next cycle building on a result that was never there.
The same caution applies to any outcome quoted as a rate or a probability. A percentage is a statement about a distribution, and the confidence interval is dropped in the summary long before it is dropped in the analysis. Before a number changes a decision, it is worth asking how many observations produced it and how wide the range around it was.
Common questions
Which single metric matters most early?
Retention within the segment you chose. Growth can be bought; retention cannot, and it is the one that predicts whether anything compounds.
When is a cohort big enough to trust?
Less a fixed number than a question of effect size: small effects need large samples. If the difference you are chasing is a few percent, the sample required is far larger than most early-stage companies have.