增长周会：渠道、漏斗、实验与下一步

1. 

   1We spend less, and installs grow

   演讲者备注

   This week, we reverse last week’s acquisition efficiency problem: we spend less and bring in more installs. In this illustrative W44 review, the recommendation is to ship the referral reward, give pricing more time to measure paid conversion, and validate checklist retention before claiming a downstream benefit. The distinction throughout is between what we observe, what we think explains it, and what the evidence supports doing next.

2. 

   2Lower spend now delivers higher acquisition volume

   演讲者备注

   Starting with that recovery, installs rise 7.0% to 41,200 while total spend falls from $88,600 to $86,100. Paid-social CPI also improves from $3.00 to $2.62, so the acquisition gain is not simply the result of buying more volume. These are week-over-week improvements, not results against a budget target. We use the workbook’s exact counts; the sync notes’ 41.2k is the same W44 result rounded.

3. 

   3Two tests finish; revenue and retention remain open

   演讲者备注

   Behind those totals, last week’s commitments turn into concrete delivery, but only some hypotheses become confirmed results. The six new creatives go live, two underdelivering influencers are cut, and the checklist ships on October 28. The pricing and referral tests reach final 14-day results with significant lifts in trial starts and invites, respectively, but pricing does not establish a revenue win—and that is not proof of no revenue effect. Paid social makes the biggest acquisition move, adding 2,300 installs without more spend, while creative fatigue and checklist retention remain open explanations or outcomes.

4. 

   4Paid social leads our acquisition recovery

   演讲者备注

   That paid-social gain is the center of the acquisition recovery: the same $48,000 now delivers 18,300 installs rather than 16,000. We can call that an efficiency improvement; we cannot yet say the creative refresh causes it.

5. 

   5Social adds volume while influencer spend falls

   演讲者备注

   Looking across the other channels helps separate where we gain volume from where we save money. Paid-social spending stays at $48,000, while influencer spending falls from $15,000 to $12,000 and influencer installs decline from 4,200 to 3,900. Referral spending rises from $3,600 to $4,100, but installs increase from 2,400 to 2,900 and CPI improves to $1.41. Search adds modest volume at unchanged spend and organic stays flat; Ravi’s proposed next step is the new “learn spanish fast” keyword set.

6. 

   6Social CPI falls after the creative refresh

   演讲者备注

   Returning to the largest gain, the creative refresh gives us a plausible explanation for social’s recovery, not a proven one. After six new UGC creatives go live on October 27, frequency falls from 3.8 to 2.7 and CPI falls from $3.00 to $2.62. That clears last week’s $2.80 aim, which was a target rather than a forecast. The timing is consistent with the fatigue hypothesis, but we should report the measured improvement without treating the coincidence as causal evidence.

7. 

   7The checklist ships; activation reaches 66.1%

   演讲者备注

   Beyond acquisition, the checklist rollout is now complete, backed by last week’s randomized activation result of 67.2% versus 60.8%. This week’s overall activation reaches 66.1%, but that weekly rate is not a new treatment-effect estimate, and retention still needs its own readout.

8. 

   8Trial starts reach 29% of activated users

   演讲者备注

   Following those users further through the funnel, we reach 5,607 trial starts out of 19,336 users who complete a first lesson, a 29.0% conversion rate. Those users sit within the W44 funnel of 41,200 installs and 29,252 sign-ups. The separate 36.2% trial-to-paid figure belongs to W42 trials, so we cannot append it to this funnel or use it to claim that the newly shipped checklist improves paid conversion. We need to keep the weekly funnel, older-cohort paid conversion, and pricing experiment results distinct.

9. 

   9D7 reaches 25%, but its cohort label needs reconciliation

   演讲者备注

   That same cohort discipline matters even more for retention, where the reported D7 rate is 25% but the underlying label conflicts. The workbook calls it the previous week’s cohort, meaning W43, while the sync notes call it W42; until that is reconciled, we cannot make a clean cohort comparison. D1 at 47% and September-cohort D30 at 12% also belong to different windows, not one retention curve. Kenji’s proposed next step is to resolve the labels and read W44 D7 by checklist exposure, since October 27 still has a 50/50 split.

10. 

    10EXP-101 moves from rollout to retention monitoring

    演讲者备注

    Once we resolve that cohort definition, the checklist’s next evidence obligation is retention, not another declaration that activation wins. Overall activation moves from 64.0% to 66.1%, but these are weekly totals rather than randomized experimental arms. The October 28 rollout closes the delivery commitment; it does not close the hypothesis that better activation improves retention. For next week, I recommend that Kenji validate the cohort and separate checklist exposure before drawing a D7 conclusion.

11. 

    11Referral earns rollout; pricing needs paid-conversion data

    演讲者备注

    With checklist retention still open, the two completed tests point to different decisions. In the pricing test’s final 14-day sample of 36,580, the $59.99 annual offer lifts trial starts to 29.4% from 27.1% at $69.99, with p=0.03. Revenue per 1,000 activated users rises from $398 to $412, but p=0.41 does not establish a revenue win, so I recommend that Sam extend measurement one week for trial-to-paid data. Referral has a clearer rollout case: invites rise from 1.18 to 1.42 with p=0.02, and referred installs rise from 7.9 to 9.6 per 1,000 users at $1.41 per install. I recommend that Lea ship the two-week reward.

12. 

    12Let's ship referral and close the revenue and retention questions

    演讲者备注

    Those different evidence standards leave us with four proposed commitments to confirm today. Lea ships the two-week referral reward, Sam extends pricing measurement for one week before a pricing decision, and Ravi launches the new search keyword set. Kenji delivers the checklist-cohort D7 readout next week, after reconciling labels and separating exposure. We finish the week with better acquisition efficiency and a shipped checklist; let’s carry that progress forward without turning unresolved revenue and retention signals into claims, and revisit these commitments before making new recommendations next week.

如何自动制作幻灯片或 PowerPoint 柔和圆角 · 13 页 季度经营回顾：业绩、驱动因素与决策 瑞士极简 · 14 页 高端酒店：旗舰物业投资提案 摄影编辑 · 15 页

查看全部案例 →