Skip to content
Back to Growth & Acquisition

Growth Automation

Automate the repetitive growth tasks (lifecycle email, churn intervention, referral attribution) so the team focuses on experiments, not ops.

Why

Growth has two types of work: experiments (high-leverage, non-repeatable, require human creativity) and ops (low-leverage, repeatable, automatable). Without automation, the team spends 60-80% of time on ops — manually sending lifecycle emails, pulling churn lists, reconciling referral attribution — and has no time for experiments. Automating the ops (lifecycle email triggered by behavior, churn intervention triggered by model score, referral attribution in code not spreadsheets) frees the team to spend 80% on experiments, which is where the learning and the growth come from. The automation also improves quality: a behavior-triggered email sends at the exact right moment, not when a human remembers, and code-based attribution doesn't have the spreadsheet formula errors that mispay rewards. The discipline is to automate every task you've done 3+ times manually, because the 4th time is a sign you've built a repeatable process.

When

When the team spends >40% of time on repeatable ops. Automate the highest-frequency ops first, then the next.

How

1. List every recurring growth task: lifecycle emails, churn interventions, referral payouts, weekly reports, experiment setup
2. Tag each by frequency (daily, weekly, monthly) and time-per-occurrence; prioritize the highest (frequency × time) for automation first
3. For lifecycle email: trigger on behavior (signup, activation, inactivity 7d) not time-since-signup; use Customer.io or Iterable
4. For churn intervention: trigger on the churn model's daily score, not a manual weekly list pull; auto-enroll high-risk users in the win-back sequence
5. For referral attribution: compute in code (the referral_attribution table) not in a spreadsheet; auto-pay rewards on activation, auto-flag fraud
6. For weekly reports: auto-generate from the data warehouse and post to Slack; no human should spend 2h every Monday on a report
7. For experiment setup: template the flag + variant + metric collection so a new experiment is 1h of config, not 1d of code
8. Measure the automation ratio: (time on experiments) / (total growth time); target 70-80%, and re-audit quarterly as new ops accumulate

Example

# Growth ops audit (before automation)
Task                    | Freq  | Time/occ | Total/wk
Lifecycle email send    | daily | 30 min   | 3.5h  ← automate first
Churn list pull + email | weekly| 2h       | 2h    ← automate
Referral reconciliation | weekly| 1.5h     | 1.5h  ← automate
Weekly growth report    | weekly| 2h       | 2h    ← automate
Experiment setup        | 3x/wk | 2h       | 6h    ← template
Total ops time: 15h/wk (of 40h = 38%) → target <12h (30%)

# After automation
Lifecycle email: triggered by behavior in Customer.io (0h manual)
Churn intervention: triggered by model score, auto-enrolls (0h manual)
Referral: computed in code, auto-pays on activation (0h manual)
Report: auto-generated, posted to Slack (0h manual)
Experiment setup: templated, 1h/config (3h/wk vs 6h)

# Result
Ops time: 15h → 3h/wk  |  Experiment time: 25h → 37h/wk
Automation ratio: 62% → 92%  (target 80% ✓)
Experiments/week: 1.5 → 3.2  (freed time went to experiments)

Common Problems

  • ⚠Automating nothing because 'it's faster to just do it manually'; the 4th manual occurrence is the sign to automate.
  • ⚠Automating before the process is stable; the automation encodes a broken process and you have to rebuild it.
  • ⚠Automating but not monitoring; the lifecycle email breaks silently and 2,000 users don't get their activation nudge.
  • ⚠Spending 2 weeks automating a task that takes 30 min/month; the ROI is negative and the team lost experiment time.

Best Practices

  • ✓Automate any task you've done 3+ times manually; the 4th time means the process is repeatable and the automation ROI is positive.
  • ✓Stabilize the process manually first (run it 3x the same way), then automate; automating a broken process creates a broken automation.
  • ✓Add monitoring to every automation (alert on failure, dashboard on throughput); a silent failure is worse than manual because no one notices.
  • ✓Prioritize by (frequency × time); a daily 30-min task is worth automating before a monthly 2h one.