Traffic tells you they came. Retention tells you they came back.
Stickiness by distinct active days, a new versus returning split per source, and named cohorts you can filter any view by. Plus a straight answer about which of it needs cookie mode.
What is retention analysis?
Retention analysis measures whether the people who visited once come back, and how often. It reframes traffic from a count of arrivals into a count of relationships: ten thousand strangers a month and two thousand people visiting five times each produce the same chart and describe completely different businesses.
It is the metric growth is actually made of. Acquisition can be bought and will stop the moment you stop paying. Retention is the part that compounds, and it is the first thing that quietly degrades when a product gets worse.
Stickiness: how many days, not just whether.
For every visitor active in the range, Mrkr counts the distinct calendar days they appeared, then buckets visitors by that count. A 30-day range can place someone anywhere from one active day up to thirty.
Two summary numbers come out of that distribution. Returning share is the proportion of visitors active on more than one day. Average active days is the mean across everyone. Watching the shape of the distribution move is more informative than either number alone, because a rise in the tail and a rise in the head mean opposite things.
What each mode can actually measure.
Cookieless mode, the default
- The visitor id is rebuilt from a salt that rotates every 24 hours.
- Within a day, sessions, journeys, funnels and revenue all work normally.
- Across days, the same person cannot be recognised, so stickiness collapses to one active day per visitor.
- No consent banner is required for it.
Cookie mode, per site
- The visitor id persists in the browser, so the same person is recognised across sessions and weeks.
- Stickiness, returning share and lifetime session counts become meaningful.
- Cohort conditions like minimum sessions and minimum revenue work over a real history.
- You do need a consent banner for that site.
Say the quiet part out loud.
Where the line actually is
Most privacy-first tools show a returning-visitor number without mentioning a rotating id cannot produce one across days.
Exact inside a single day
Sessions, entry and exit pages, funnels, campaign attribution and revenue reported on the visit are all exact cookielessly.
Cookie mode is a switch, not an upgrade
Per site, reversible, on every plan. It buys real multi-day retention and costs the no-banner claim for that site.
Three retention questions, three answers.
Mrkr does not ship a single retention chart. It ships three different measures, because the word covers three different questions and one chart answers only one of them.
| Question | Where Mrkr answers it | Needs cookie mode? |
|---|---|---|
| Did they come back at all | New versus returning, split by source | Yes, for a meaningful split |
| How often do they come back | Stickiness, distinct active days per visitor | Yes |
| Who exactly keeps coming back | Cohorts, applied as a filter on any view | For history-based conditions |
| Did this campaign bring repeat visitors | New versus returning by source | Yes |
| Are they going deeper each time | Engagement: scroll depth and engaged time | No |
| What share of each signup week returned in week 4 | Not available in Mrkr | n/a |
The last row is the honest one. There is no acquisition-cohort retention triangle in the product. If that specific chart is a requirement, a product analytics tool built on a columnar store is a better fit and you should not let anyone talk you out of it.
See cohorts and engagement
The live demo has saved cohorts, a stickiness distribution and scroll depth on real demo traffic. No signup, no card.
Open the live demo
Reading the numbers without fooling yourself.
A short range flatters you
Nobody can have six active days in a five-day window. Compare 30 days against 30 days, never against last week.
Returning share moves with acquisition
A good campaign floods the range with first-timers and drives the share down while the business improves. Read the count too.
Bots would ruin this metric
Automated traffic is the most regular visitor you will ever have. Nine filter layers keep it out of the denominator.
One number is not a trend
Stickiness is noisy on small sites. Annotate releases and campaigns so you compare against a labelled timeline.
Questions, answered.
Keep reading
- Cohort analysis
Naming a group of visitors once, then filtering every view by it.
- Engagement metrics
Scroll depth, engaged time, and the stickiness distribution in full.
- Engagement documentation
The exact stickiness definition and an honest read of the denominator.
- Website visitor tracking
What identity means in each mode, and what it costs either way.
- Funnel analysis
The other half of the picture: whether they finish what they started.
- Every Mrkr feature
The full index of what the product measures.
Your first visitor is already here.
Drop in the script and watch them land. It takes about a minute.