Read your first cohort chart: the no-floor week-8 retention test

You have a small group of real users, a spreadsheet, and a retention chart that keeps falling. Expect the first drop. A no-floor decay curve warns that the curve never stops falling.

By early 2025, the startup was down to roughly six months of runway, and its forecast suggested only a breakthrough would save it. It projected cash exhaustion in Q1 2026 and reached that date almost to the week. For a small team, that chart was the earliest honest forecast available.

The chart predicts shutdown

With no distribution yet, retention gives the only forecast that does not depend on a paid channel, a viral loop, or a big launch. It tells you whether the people you helped by hand are coming back, or whether you are polishing a door nobody re-enters. Growth tactics that need scale to work are useless here. You need a diagnostic that works on the first users you actually talked to. The retention line is the first number you can trust before you spend on paid growth.

Start with the users you already have. Manual outreach gives you a list of names, not a funnel. Use that list. Ask each person what they did after the initial session, what they would do again, and what made them stop. Those answers feed the chart. Those interviews cost less than any test you will ever run.

Build the no-floor diagnostic before the cash runs out

Do this before you add any acquisition channel, because returning users are the real signal. A new user who returns outperforms a crowd who never comes back again. Keep the diagnostic as a small table, not a blended average: each signup group gets its own early-return lines and its own week-8 line.

• Track day-2, day-3, and day-7 returns as separate cohorts. The first return gives the clearest test of whether the product did something worth repeating. For the January group, 57% came back on day 2, but by April the return rate had dropped to 31%. When the early return line falls, check the first moment of value before you touch the email copy.
• Compare the first cohorts against each other. A one-time spike can hide a weakening product. In the January cohort, 74% were active in week 1, 31% in week 4, and 9% in week 8, while the February cohort was weaker: 68% in week 1, 24% in week 4, and 5% in week 8. Put the lines on the same chart so the gap is visible without a tooltip. If the newer cohort is weaker, your onboarding may be decaying. If it is stronger, you may be finding a better fit, but the floor still has to prove it. Fix the gap before you add more users.
• Plot the curve on a scale that shows the bend. A flat line can hide a gradual decline. When a retention curve flattens into what is sometimes called a smile curve, it settles above zero, for example at 15%, 25%, or 40%, suggesting a durable group of users keeps returning. The chart works when the line stops falling and holds a floor, rather than sliding toward nothing. If it never levels, the first users are not finding a reason to come back.
• Set the week-8 line as the pivot trigger. Early enthusiasm does not build a business. The rule of thumb says week-8 active retention below 20% indicates the need for a different product or audience. Make the rule boring: a red line on the chart and a written decision before the next cohort arrives. Once the line is crossed, stop adding features and pivot the offer, the audience, or the core workflow. Put the rule in the same place as the chart.
• Use the signal before the cash runs out. A chart costs less than a shutdown. A company saw week-8 retention stay below 10% across five straight cohorts. That pattern marks the moment to stop optimizing onboarding and ask whether the core job is real. A wrong product calls for a different core workflow. A wrong audience calls for a different person to talk to. The memo should name the curve, the missing value, and the next test.