At a glance
Cohort Retention D7 % is the share of a signup or first-action cohort that returns and performs a qualifying event seven days after joining. It is the earliest strong retention signal you can read, much faster to act on than D30, and it tells you whether your onboarding and first-week experience actually earned a second visit. In Mixpanel terms, this is a retention report grouped by acquisition cohort, read at the day-7 column. Because it moves quickly, a fall in D7 is one of the first symptoms of a broken onboarding flow, a bad campaign cohort, or a product change that hurt the first-week experience. Watching D7 lets you correct course weeks before the damage shows up in D30 or in revenue.
| What it counts | The percentage of users in a defined cohort (typically grouped by first event or signup date) who fired a qualifying return event on day 7 after joining. |
| Sample type | Backend API data from Mixpanel retention reports, computed per cohort against the project’s event stream. |
| Why it matters | D7 is the fastest reliable read on whether new users stick. It surfaces onboarding and first-week problems while you can still fix them for the next cohort. |
| Reading the value | A higher percentage is better. A sudden fall versus the recent baseline is the headline signal, treat a drop of more than a few points as worth investigating. |
| Currency | percent |
| Time window | 30D vsP |
| Alert trigger | drop >5pp vs baseline |
| Sentiment key | mix_cohort_retention_d7 |
| Roles | owner, marketing |
Calculation
Vortex IQ reads your Mixpanel retention report grouped by acquisition cohort and takes the day-7 retention column. For each cohort, the numerator is the number of users who fired the qualifying return event on day 7 after their first event, and the denominator is the original cohort size. The card surfaces the blended D7 percentage across recent cohorts over the selected window and compares it against the recent baseline. The alert latches when the current reading falls more than five percentage points below that baseline. Because D7 only needs a cohort to age seven days, it updates far sooner than longer-horizon retention numbers.Worked example
A representative reading of Cohort Retention D7 % for a typical merchant on Mixpanel. Suppose 800 shoppers signed up or placed a first order on 14 Jun 26. Seven days later, on 21 Jun 26, you check how many returned and fired a qualifying event such as a repeat session. If 312 came back, your D7 reads 39%. Your recent baseline across the last few cohorts was around 40%, so this cohort is healthy and no alert fires. Now imagine the next cohort reads 32% after a checkout redesign shipped, that is an 8pp fall against baseline, so the alert latches. You open the retention report, confirm the dip is concentrated in users who hit the new checkout, and roll back or patch the flow before the following cohort is affected. For deeper investigation, use Vortex Mind to trace upstream causes; for natural-language exploration, ask Ask Viq.Sibling cards merchants should reference together
| Card | Why merchants reach for it |
|---|---|
mix_cohort_retention_d30 | The longer-horizon checkpoint. Compare D7 to D30 to see whether early retention holds. |
mix_retention_curve | Shows the full decay curve so you can see how steeply users fall off after week one. |
mix_worst_decaying_cohort | Pinpoints which acquisition cohort is losing users fastest, often a specific campaign. |
mix_stickiness | DAU/MAU stickiness is the habit signal that drives healthy week-one retention. |
mix_new_users | New vs returning users frames how much of your base is first-time versus repeat. |
Reconciling against Mixpanel
Where to look in Mixpanel’s own dashboard: Open your saved Retention report, set the birth event to the same first action the card uses and the return event to the same qualifying event, then read the day-7 column for the cohorts in your window. The blended figure on the card should track the average of those cohort rows. Confirm the retention type (recurring vs unbounded) matches, because the two compute the day-7 column differently. Why the Vortex IQ value may legitimately differ:| Reason | Direction | What to do |
|---|---|---|
| Cohort maturity. Cohorts younger than seven days cannot yet report a D7 value. | Vortex IQ excludes immature cohorts from the blend | Compare only cohorts that have aged seven full days in both views. |
| Retention type. Recurring vs unbounded retention count the day-7 return differently. | Variable | Match the retention type used by the card. |
| Event definition. The birth or return event differs between the report and the card. | Variable | Align both event definitions before comparing. |