Where did first-time buyers stop returning?
I broke a repeat-order goal into earlier milestones and channel differences, helping lifecycle operations focus on the first useful return.
- My role
- Lifecycle diagnosis & cadence design
- Original tools
- SQL · Excel · BI
- Deliverable
- Channel-specific follow-up priorities
The problem
Customers who reached four orders within 60 days had stronger retention, but few reached that point. Which earlier step needed attention?
What I did
- Separated first-to-second, second-to-third, and third-to-fourth order progression by acquisition channel.
- Used repeat-order timing and test review to propose earlier follow-up, including a second-order goal within 30 days.
Store-acquired customers struggled most with the first repeat order.
Only 16.7% of store-acquired first-order customers reached a second order within 60 days, compared with 50.0% from app discovery. Later transitions were less useful for locating the channel gap in this reconstruction.
Locate the gap using each transition’s own denominator
View data & definitions
Channel groups are mutually exclusive in this example. The same 60-day observation horizon is used for all nested order milestones. Similar later-stage rates illustrate the first-repeat bottleneck; they are not a claim of exact equality in the company data or equal treatment effects.
| Channel | First-order cohort | Second by day 60 | Third by day 60 | Fourth by day 60 | Second by day 30 |
|---|---|---|---|---|---|
| Store acquisition | 120,000 | 20,000 | 12,000 | 6,000 | 16,000 |
| App discovery | 80,000 | 40,000 | 24,000 | 12,000 | 28,000 |
| Partner referral | 60,000 | 26,000 | 15,600 | 7,800 | 19,000 |
Decision: Give the first repeat a channel-specific diagnosis, including whether store-acquired shoppers know other stores and categories are available.
Order timing helped turn a broad goal into earlier checks.
Among customers who eventually reached four orders, repeat orders followed a recognizable cadence. Those patterns provided a monitoring reference, while leaving open what would help slower or non-returning customers.
Timing among customers who reached four orders
View data & definitions
Percentiles summarize days since first order among eventual fourth-order completers only. They do not estimate all-customer conversion or prove that imposing an order deadline improves retention. The historical work used cadence and push-test review to guide operational timing.
| Completer percentile | Second-order day | Third-order day | Fourth-order day |
|---|---|---|---|
| 50.0% | 9 | 21 | 33 |
| 65.0% | 12 | 27 | 40 |
| 80.0% | 17 | 35 | 49 |
| 90.0% | 22 | 41 | 56 |
Use milestones as prompts to investigate a customer need.
- Second order within 30 days.
Prioritize the first repeat, especially for store-acquired customers.
- Review progress around day 40.
Look for customers who have not moved toward a third order and inspect the barrier.
- Review the 60-day outcome.
Track fourth-order attainment alongside customer value, cost, and later retention.
These are monitoring proposals. Reaching a milestone is associated with retention; the analysis does not establish that pushing everyone to the same order count causes it.
Focus the response on the next useful order.
What I delivered: A channel-specific lifecycle diagnosis and follow-up priorities, supported by order timing and historical test review. The second-order goal brought an earlier, actionable checkpoint into the longer repeat-purchase agenda.
Cohort definitions, timing limits & source notes
Based on my JD Daojia Q3 repeat-order project and operations-sharing material. Original analysis used SQL, Excel, and BI; the replacement cohorts and Python charts are portfolio reconstructions.
- All channel cohorts have 60-day follow-up. Second, third and fourth order within that horizon are nested counts. Second order within 30 days is a separate goal; it is not required for every fourth-order completer.
- Conditional transition rates use different preceding-stage populations. Channel differences are descriptive and may reflect customer mix and purchase context.
- Timing percentiles condition on reaching four orders by day 60. They cannot estimate what non-completers would have done or the effect of an intervention.
- The historical deck reports push-test improvement, but its comparison design and denominator are not sufficiently specified here to reproduce a causal lift figure.
Original decks, channel identifiers, customer records, operational rules and SQL remain private.
Download milestone and timing data