← All guides

Best practices

Adding Metrics That Matter

A number does something to a sentence that no adjective can. "Significantly improved processing time" asks the reader to trust you. "Cut processing time from three days to four hours" gives them nothing to doubt.

The right metric depends on the work

• Revenue roles have it easiest: dollar amounts, growth percentages, deal sizes, win rates, conversion
• Operations speaks in time and money saved, error rates, throughput, cycle time
• Technical work has latency, uptime, deployment frequency, and scale, where 10 million events a day says more than any framework name
• Leadership is measured in people: team size, retention, how many of your reports got promoted
• Project work lives on delivery: on time, on budget, adopted by how many
• Support and success run on volume, resolution time, deflection, retention, and CSAT

Three ways to build a number when you have none

The common objection is "I don't have exact numbers." You almost always have enough to get to one honestly, by one of three routes.

Arithmetic is the first. If you automated a 30-minute daily task for a team of ten, that is about 1,250 hours a year, and the arithmetic itself is the credibility. You can walk anyone through it.

Before-and-after is the second, and it is often the most persuasive, because it carries its own context. "Reduced onboarding from 14 steps to 5" needs no baseline explained; the reader supplies the judgment themselves.

Ranges are the third. "Between 15 and 20 hours a week" is honest, defensible, and infinitely better than dropping the claim because you cannot pin it to a decimal.

Context is the other half of every metric

Half a million dollars in revenue is a company-changing number at a seed startup and a rounding error at a Fortune 500, so say which world you were in. Team size, customer count, market, and timeframe all do the same work, and each of them costs you about four words.

The metrics that quietly impress

Most people reach for the biggest number they can find. Experienced readers are usually more moved by two other kinds.

Durability is the first: a result that held. "Held churn under 4% for six consecutive quarters" is harder to achieve, and harder to fake, than a single good month. Second-order effects are the other: what the result made possible. A migration that cut latency 70% is a fact; a migration that cut latency 70% and made real-time analytics viable for the first time is a business event.

Where numbers become a liability

Never publish a number you cannot defend out loud. Hiring managers cross-reference, and interviews are where inflated metrics go to die. The standard is not impressive, it is defensible. An honest 15% told well beats a suspicious 300% every time.

Two specific traps are worth naming. Percentages without a base invite the obvious question. A 300% increase from two customers to eight is technically true and reads as evasion once the interviewer does the arithmetic, so give the base yourself. And claiming a team result in the first person singular is the fastest way to lose a room; "I grew revenue 40%" collapses the moment someone asks who else was on the account. Write "contributed to" or name your part, and the whole bullet gets stronger rather than weaker.