Fyxer says 90% of users are still active after three months. What an FDE should ask before believing it

Fyxer, a startup that makes an AI email assistant, says 90% of its users are still active after three months. The figure appears in the company’s own Series B funding announcement, where it is offered as evidence that the product gives users a lot of value.

Unite.AI repeated the figure and called it a rare level of retention for productivity software. Silicon Canals also ran it, but made clear that it came from the company and did not say that anyone had verified it independently.

This matters to FDEs. In client meetings you will often hear lines like “90% of users stick around”, sometimes in a vendor’s materials and sometimes in your own team’s slides. Clients trust the person who can tell a defined metric from a marketing line.

What is missing from the number?

Fyxer’s announcement does not define a cohort or describe how the figure was measured. According to Unite.AI, Fyxer serves more than 180,000 users, but no source says whether the 90% covers all of them or only a group that signed up in a particular month.

Fyxer’s post also offers a 14-day free trial. Are trial users who never convert to paid included in the denominator? If not, the 90% measures how many people stayed among those who had already decided to pay, a group that has been filtered in advance.

Same data, two retention figures

Take a hypothetical cohort, not Fyxer’s data: 1,000 people sign up in the same month. 400 use the 14-day trial and leave, 600 convert to paid, and after three months 540 are still active.

If the denominator is paying users only, retention is 540 divided by 600: exactly 90%. If the denominator is everyone who signed up, retention is 540 divided by 1,000: 54%. Not a single row of data has changed. Only the definition has.

In SQL, that difference often comes down to one WHERE clause that filters out trial users. So when you calculate retention yourself, run both versions and write the definition next to the number.

Even the revenue figure varies

Small discrepancies between reports show how easily numbers shift as they are copied from one place to another. Fyxer says its run-rate revenue grew from $1 million at the start of 2025 to more than $17 million by August of that year, and Unite.AI reported exactly that. Silicon Canals reported more than $16 million.

The gap is small, but it is a useful lesson. When two sources disagree, go back to the original text and check how the metric is defined. Run-rate revenue and ARR are not necessarily the same thing, just as “active” at one company may not mean the same as “active” at another.

Fyxer has about 48 staff. The round was led by Madrona, with Lakestar participating. The question here is not whether Fyxer is right or wrong. It is the habit of repeating figures that startups publish about themselves without the definitions that go with them.

What should an FDE do with numbers like these?

If a client sends you a similar figure to justify buying a tool, do not dismiss it. Ask questions instead. That keeps the conversation about data rather than impressions.

Ask who measured it, whether the denominator is all users or a single cohort, and what counts as active: opening the app, sending one email, or using the core feature. Ask too how trial users are counted, and whether there is an original source to check against.

For developers moving into FDE work, this is a skill you can practise in your current job. Write a cohort retention query on your own product and record the definition alongside it. If the job descriptions you are targeting mention adoption or success metrics, that is where to bring this skill up in interviews.

On a CV, a line such as “redefined the active-user metric for a client dashboard” is worth more than a percentage with no context. Fyxer may well keep 90% of its users. But a good FDE only puts that number in front of a client after finding out how it was measured.