Bookwise Social Activation

Page 7 of 7 — So what, and what next: implications of the magic number

The Bookwise Activation Metric

A user is socially activated when they follow ≥5 people within 7 days of signup.

45%

W12 retention (activated)

13%

W12 retention (not activated)

32pp

Retention gap

39pp

Gap (organic users only)

Why ≥5? This threshold was derived from retention curves (see page 6), following the Facebook “7 friends in 10 days” methodology. At ≥5 following in 7 days, the retention curve transitions from declining to stable: 45% of users are still active at week 12 vs 13% for those below the threshold. Among organic (non-December) users, the gap widens to 59% vs 20% (39pp). Lower thresholds (≥1, ≥3) show retention gaps too, but ≥5 is where the curve inflects — each additional follow below 5 adds 3–4pp of retention, while above 5 the marginal gain drops to 1–3pp with rapidly shrinking populations.

Retention Separation: Activated vs. Not

Users who follow ≥5 people in their first week are 3.5x more likely to still be reading at month 3. The retention curves (see page 6) show the gap opens immediately in week 1 and remains stable through month 4.

3.5x

retention ratio at week 12 (45% vs 13%)

Effect Size Comparison Across All Hypotheses

The Mann-Whitney hypothesis tests (page 4) provide supporting evidence, but the activation metric was derived from retention curves (page 6) — the Facebook-style methodology of finding where cohort survival rates separate. The effect sizes below confirm that following is the strongest actionable lever; the retention curves show why ≥5 is the right threshold (32pp gap at week 12, 39pp for organic users).

H6 Momentum

0.805

H3a Kudos Recv

0.734

H3b Kudos Given

0.683

H4 Combined

0.623

H5 Bidirectional

0.506

H1 Following ≥5

0.467

H2 Followers

0.413

Note: The effect size chart above is one lens. The retention curves on page 6 are the primary evidence: they show the ≥5 threshold is where the curve transitions from declining to stable. Hypothesis test effect sizes confirm the signal is real; the curves tell us where the threshold lives. Kudos interventions are already proven at Bookwise — finish-book posts trigger push notifications to followers, who tap through and give kudos. This existing loop can be amplified. Only 1 out of 212 active readers has zero social connections, confirming that social engagement is near-universal among retained users.

✦

The Confound: Reading Predicts Reading

H6 (reading momentum) has the highest effect size at 0.805 — users who read in their first week continue reading. This is the baseline we must account for.

The stratified analysis (page 5) addresses this directly: among users who read the same amount in their first week, those who followed ≥3 people read 2.3x more in the following month. This holds specifically for mid-range readers (1–15 sessions) — the population where social engagement has independent value controlling for past reading. Social metrics are valuable because they predict reading above the momentum baseline, and because they represent levers we can actually pull. We can’t force someone to read, but we can help them follow 5 people in their first week.

Why ≥5 Following in 7 Days?

The retention curves on page 6 are the primary evidence. Following the Facebook methodology of plotting cohort survival by behavioral threshold, ≥5 is the inflection point where retention transitions from declining to stable. Hypothesis tests (page 4) and the stratified analysis (page 5) provide supporting evidence that the signal is real and independent of reading momentum.

Retention curve inflection point

The retention curves show ≥5 is where the step change happens: 45% week-12 retention vs 13% below the threshold (32pp gap). For organic users, the gap is even starker: 59% vs 20% (39pp). Thresholds above ≥5 continue to improve but with diminishing returns and shrinking populations.

Meaningful commitment threshold

Following 1–2 people might be accidental. Following 5 means the user has actively built a reading circle. 33% of users already hit this threshold organically (151 of 452), confirming it’s achievable but not trivial.

Clear intervention path

Every hackathon experiment (A2–A5) is designed to put followable people in front of new users. If we can get a new user to follow 5 people in their first week, the retention curves predict they’re 3.5x more likely to be reading at month 3.

Simple to measure

One number, evaluated at day 7: how many people does this user follow? No complex aggregation, no dependency on other users’ behavior (unlike follower count or bidirectional metrics). The metric comparison table on page 6 confirms following has the largest retention gap of any actionable metric.

Implications for Hackathon Experiments

The question for each experiment: does this help users follow 5 people in their first week? The activation metric tells us what predicts retention; only experiments will tell us if we can create it. Each experiment below is evaluated by how directly it drives the ≥5 target.

A2 — Direct

Same-Book Connection

Directly drives follows toward the ≥5 target. Each same-book match that converts to a follow is a step toward activation. This is the most direct path to the magic number.

Targets: following ≥5 (primary activation lever)

A3 — Direct

Reading Neighbors

Same mechanism as A2 — directly drives follows toward the ≥5 target — but with a broader taste-matching net. Good for users without obvious same-book connections who still need 2–3 more follows to hit the threshold.

Targets: following ≥5 (wider matching net)

A4 — Moderate

Flyby (Lightweight Connection)

Lighter touch — surfaces coincidental reading overlap. May contribute 1–2 follows toward the ≥5 threshold, but unlikely to drive activation on its own. Best as a supplement to A2/A3.

Targets: following (contributing 1–2 toward ≥5)

A5 — Indirect

Hot Takes

Indirect path to the magic number. Drives engagement (kudos, comments) that may lead to profile discovery and follows, but the chain from engagement to ≥5 following is longer and less certain.

Targets: engagement → discovery → follows (indirect)

✦

Limitations

i.

Correlation, not causation

Both social engagement and reading may be driven by a third factor like intrinsic motivation or pre-existing reading habits. We cannot conclude that creating social connections will cause more reading.
ii.

Small sample

460 clean users, of which only 212 are active readers. Effect sizes are large, but confidence intervals are wide. A 10x larger cohort would be needed for reliable subgroup analysis.
iii.

Survivorship bias

We only observe users who stayed long enough to generate data. Users who churned before any social activity are invisible to this analysis, yet they are exactly the population we most need to understand.
iv.

Predictive, not prescriptive

The activation metric tells us what predicts retention. It does not tell us whether our interventions can create the conditions it measures. The experiments will test that.

Recommended Round 2 Analysis

The retention curves answered “what predicts retention” and identified the magic number. These next steps address whether we can trust it across populations and whether we can move it.

1

Track the magic number across new cohorts

The cohort sensitivity analysis on page 6 revealed that the Dec 2025 Dragonsteel Nexus cohort (198 users, 43% of the dataset) behaves very differently: only a 16pp retention gap vs 39pp for organic users. Future analysis should track whether the ≥5 threshold holds for new signup cohorts, or whether event-driven cohorts need a different activation model entirely.
2

Disentangle social engagement from reading momentum

The stratified analysis (page 5) already shows that social engagement has independent value (2.3x lift controlling for past reading among mid-range readers). A multivariate model could further isolate the independent contribution of following: does hitting ≥5 follows predict retention beyond what first-week reading sessions alone explain?
3

Test interventions experimentally

The activation metric tells us what predicts retention, but only experiments (A2–A5) will tell us if we can create activation. The hackathon experiments are the next step — ship the interventions, measure whether they move the “follow ≥5 in 7 days” metric, and see if the retention lift follows.

✦ ✦ ✦

Bottom Line

5 follows in 7 days is the Bookwise activation metric. Users who hit this threshold retain at 45% at week 12 vs 13% for those who don’t (32pp gap; among organic users: 59% vs 20%, 39pp gap). The stratified analysis confirms social engagement has independent value: 2.3x reading lift controlling for past reading among mid-range readers. The A2 (same-book) and A3 (reading neighbors) experiments are the most direct paths to helping users hit the magic number. Every experiment, every notification, every feed injection should be measured against this: did we help the user follow 5 people in their first week?