Illustrative expense workflow: each gate has an event, and the first approved report marks the chosen outcome.
- 1↓
Account created
The cohort and activation window begin.
- 2↓
Employee invited
Someone can now submit a report.
- 3↓
Report submitted
The form enters approval.
- 4↓
Report approved
The first result under this definition.
- 5
Next cycle
Check repeat use only after the full window.
Choose an outcome the product can verify
Consider an illustrative B2B expense workflow. An employee submits an expense report and a manager approves it. The product team wants to know whether new accounts have put that workflow to use. Sign-up, an employee invitation, and a submitted form show progress. For this example, the first approved report is the endpoint. Payment happens later and is outside this activation measure.
Count company accounts, because several employees may contribute to one rollout. The approval event needs an account ID, timestamp, and a status that means approval actually happened. A view of the approval screen or a request added to a queue does not qualify. Read a few event records against the underlying workflow before putting the rate on a dashboard.
A compact measurement card
The windows below belong to the illustrative workflow. They are not B2B SaaS standards; set your own from the pace of the customer job. In each row, the numerator must be a subset of the stated denominator.
| Measure | Numerator | Denominator | Window |
|---|---|---|---|
| Activation | Accounts with a first approved report | All new accounts in the chosen cohort | 14 days after account creation |
| Submission step | Accounts in that group with a first submission after the invitation | Cohort accounts that invited an employee within the window | The same 14 days after account creation |
| Repeat use | Accounts in that group with another approval within 30 days of the first | Accounts whose first approval was at least 30 days ago | 30 days after the first approval |
Read the rate alongside the people it leaves out
Here is one hypothetical calculation. Forty accounts sign up in the same week, and every account has now completed a 14-day observation window. Twelve get a first report approved within that window. Activation is 12 / 40 = 30%. This is arithmetic for a fictional cohort, not a benchmark or a product result. Add accounts from yesterday and the denominator will describe a different opportunity to finish.
For time to value, measure account creation to first approval for those twelve accounts. Their median or p75 describes completers only. The other 28 are absent from this time calculation, so put “12 of 40 within 14 days” beside it. Otherwise a few fast customers can make the whole cohort appear to have started successfully.
A funnel tells the team where to investigate. Are invited employees failing to submit, or are submitted reports waiting for a manager? Join those roles by account rather than drawing them as one person's path. If the definition of approved changes, recalculate earlier cohorts where possible or start a new series. Otherwise a tracking change can masquerade as a change in customer behavior.
Investigate the reason behind a drop-off
A delay between submission and approval could come from an absent manager, missing receipts, or a failed notification. The funnel cannot distinguish them. Review a few delayed accounts: what state was the report in, who received the request, and what did the employee see? The useful fix depends on the answer.
Accounts that saw a tip may approve reports more often, but that association does not show the tip caused the difference. The product may display it only to people already submitting a report. For a causal claim, define eligible accounts, the endpoint, and the window before randomly assigning accounts to guidance and control. If there are few customers, studying the workflow may be more useful than drawing a firm conclusion from a noisy percentage.
