Journal · 11 min · Helen Marlowe
Reading retention without the usual traps
A retention curve is a sentence with the verbs missing. Most App Analytics tools will draw the line; almost none will force you to write what “came back” meant that week. In the Product Telemetry Studio we treat the chart as a draft, not a verdict.
Pick a window you can defend in a corridor
Seven days is a habit, not a law. If your value moment is a weekly shop, D7 will flatter browsers and punish people who intended to return on day eight. If your value moment is a same-day booking, D7 is a eulogy. Write the window next to the chart, including the timezone you used. If you cannot say why D14 would be wrong, you are not finished.
We ask students to keep a short appendix: three rejected windows and a reason each. It looks pedantic. It stops a later rebrand from quietly moving the goalposts.
Returning is not the same as resurrected
A user who opens the app after 40 quiet days is not “retained” in the same way as someone who never left. Blend them and you will brief leadership on a bounce that is actually a win-back campaign, a billing reminder, or a push you should be ashamed of. Split at least once: still-active versus returned-from-silence. If the tool cannot split, say so in the brief rather than drawing a smoother line.
Label the dip
Unlabelled dips get filled with folklore. In one consumer finance case we later wrote up on the Voices page, a gentle D7 was mostly a billing artefact. The useful work was not a new colour on the chart. It was a sentence: “this curve includes people whose card was charged and who opened a receipt.” Once labelled, the dip on D9 stopped looking like a product failure and started looking like a receipt no one needed to open twice.
If you cannot label a dip after an hour with engineering, do not present the curve. Present the gap. That is still App Analytics. It is simply honest.