Your CFO doesn’t want a list of projects. They want to know what they’d lose if the three of you left, and whether the team gets more done next quarter than this one. Bullet points can’t answer either question.
DataKitchen’s answer, from the two downs make two ups post, is four metrics. Two go down. Two go up because the first two did.
The one-page report
| Metric | What you count | Where the number comes from | Direction |
|---|---|---|---|
| Errors | Defects that reached a customer last quarter, and who found each one | Incident tickets, test failures, the Slack thread where finance asked why revenue moved | Down |
| Cycle time | Days from an accepted request to a result a user can act on | Ticket timestamps, deploy logs | Down |
| Productivity | Work delivered per sprint, against your own baseline | Story points or completed requests | Up |
| Customer satisfaction | A 1-to-10 score and comments from the people who use your data | A short survey each quarter | Up |
Four numbers fit on one page. That’s why a team keeps them up to date.
Start with errors. The ostrich problem post suggests counting how many of last quarter’s incidents someone outside the team found first. That number is usually uncomfortable, and it’s the one a CFO reads without a glossary. Measure the team on defects that reach customers, not defects you find yourselves, so catching your own problems counts as a win. The shame game post explains why that matters: if reporting an error means owning the fix, nobody looks.
Put a price on the waste
Finance thinks in payroll. So translate.
The Business Case for DataOps cites a March 2020 Gartner survey in which data professionals spent 56% of their time on operational execution and only 22% on innovation that delivers value. That post’s method: test writing and automation cost about 20% of each person’s time, and it assumes they eliminate the 56%. Net, 36% of capacity returns to new work.
Here is that arithmetic for a team of three. It’s illustrative. Swap in your own salary figure and your own time split.
| Line | Figure | How it’s computed |
|---|---|---|
| Full-time employee cost per person | $156,000 | The 2021 post’s example figure |
| Team cost | $468,000 a year | 3 × $156,000 |
| Time on operational work (56%) | $262,080 a year | 56% × $468,000 |
| Net capacity returned (36%) | $168,480 a year | 36% × $468,000 |
| In headcount | About 1.08 people | 36% × 3 |
So the pitch to a CFO is a hire you don’t make. “Give us the tooling and we’ll deliver what four people would” is a sentence finance can check next quarter, against the same four metrics.
Don’t stop at the cost side. The tool is part of the math too. At October 2026 list prices on the pricing page, TestGen Enterprise for three users on one database connection is four units at $100 a month, or $4,800 a year. Open-source TestGen is $0 for one user and one connection. DataKitchen makes TestGen, so check that against the full cost comparison.
Why errors carry the dollar figure
An error that reaches a customer costs far more than one caught early. The 1:10:100 rule, from quality experts George Labovitz and Yu Sang Chang, prices a defect at 1 to prevent at the source, 10 to remediate after creation, and 100 once it reaches a customer. Shift-left testing is the practice built on that rule.
For a CFO, that turns your error count into a cost trend. Fewer customer-found errors this quarter than last means the expensive column shrank.
A benchmark helps. In DataKitchen’s 2024 post on observability results, one customer ran 70 data sets and billions of rows a month with less than one error per quarter. Your team won’t start there. You only need the trend line pointing at it.
Cycle time is the number nobody has
Most teams can’t say how long a request takes, because the clock they track starts when coding starts. Start it when the request is accepted. The waiting in between is the cost. In a 2020 DataKitchen post on analytic cycle time, a large telecom company took more than four months to release features through a four-stage manual deployment, and an Eckerson survey it cites found 76 percent of organizations take weeks or months to deploy.
A three-person team that turns a request around in days has a story to tell. Write down the number now, so you have a baseline when you do.