Sourced analysis · Documented facts are linked to their primary references. Interpretation is identified in the text.
Choose the cohort before the campaign launches
An install is an acquisition event. It cannot tell a producer whether a viewer reached the second episode, returned after a lock, or discovered a second title. Build one cohort table per acquisition week: newly acquired unique accounts, accounts that started episode one, accounts that reached the chosen story milestone, and accounts active on day 1, day 7 and day 28. Keep paid, owned and partner referrals in separate rows. The point is not to force every platform into the same metric; it is to make the denominator visible.
Define “active” before anyone sees the result. For a fictional serial, Glass Harbour, the team might define it as a logged-in account that starts any episode, not merely opens the app. A cohort of 10,000 newly acquired unique accounts produces 6,000 accounts that start, 2,400 accounts that reach episode five, 1,800 day-7 active accounts and 720 day-28 active accounts. The useful figures are 60% start rate, 40% reach episode five among starters, 18% day-7 retention and 7.2% day-28 retention, each using the same acquired-account cohort. None is “the fan count.”
Add a story milestone and an offer milestone
A general day-7 rate can conceal a serial problem. Add a milestone that the creative team can recognise: the reveal at episode eight, the first paid gate, or the point at which a lead changes allegiance. For Glass Harbour, tag episode eight as the “identity reveal.” If 2,100 accounts reach it and only 900 are active the next day, the question is whether the ending, gate, notification or audience mix changed—not whether the premise is universally weak.
Record offer exposure separately: viewers shown a subscription screen, viewers shown a coin offer, viewers who watched an ad, and viewers who purchased. Store-listing language can establish that an app has in-app purchases or a subscription path, but it cannot establish a universal live offer or a conversion rate. ReelShort’s US listing itself says its weekly price may vary by region. [Apple App Store / NewLeaf Publishing] Never infer price sensitivity from a cohort unless the exact offer, territory, device and observation date are attached.
Read movement as a question, not a verdict
Compare like with like: the same acquisition channel, territory, account-age rule and observation window. A trailer campaign that brings in 30,000 low-intent newly acquired accounts can depress day-7 retention even if it increases total episode starts. Conversely, a narrow creator referral can look unusually loyal because its audience already understands the premise. Put both acquisition volume and downstream behaviour on the worksheet.
The next meeting should end with one testable change. Example: retain the Glass Harbour first scene, but replace the campaign clip’s misleading confrontation with the actual mystery promise; then compare the share of accounts reaching episode five across two labelled cohorts. Do not call a cohort “organic” just because a paid touch was not recorded, and do not merge re-installs with new accounts without saying so. The reader’s job is to protect a useful signal from convenient arithmetic.
Sources & evidence
Apple App Store / NewLeaf Publishing · Store listing
Source date: Not stated by source · Checked: 19 Sept 2026
- The US listing identifies NewLeaf Publishing and says the app is free with in-app purchases.
- It describes a weekly subscription while stating that actual price may vary by region.