返回提示词目录第 1161 页

分析客户复购率与再次下单周期

基于订单数据计算固定观察期复购率与下单间隔,比较客户分组差异,输出运营实验方案及代码。

完整提示词

<context>
You are a retail analyst. In most shops the biggest drop is between the first and second order: once a customer buys twice, the chance of a third rises sharply. So the second order is the number to understand. Two traps distort it. Customers who joined recently have not had time to buy again, so a raw "percentage who reordered" mixes old and new customers and falls whenever acquisition grows; the fix is a fixed window (repeat within 90 or 180 days) measured only on customers old enough to have had the full window. And a heavy first-order discount can attract one-time deal hunters, which shows up as a lower repeat rate for that code or channel.
</context>

<task>
Analyse repeat purchasing in this data.

<order_data>
{{order_data}}
</order_data>

Product context: {{product_context}}

1. Data checks: guest checkouts that split one person into several ids, cancelled or fully refunded orders (exclude), marketplace orders without customer ids, multiple orders on the same day (merge or not, say which).
2. Repeat rate: share of customers who place a second order within 30, 90 and 180 days of their first, each measured only on customers whose first order is at least that many days before the data end. Show it by first-order month to see the trend.
3. Time to second order: median and the distribution (share by week or month bucket); compare with the natural replenishment cycle given in the product context.
4. Reorder intervals: median days between consecutive orders for customers with 3 or more orders, and how the share of customers continuing falls from order 2 to 3 to 4.
5. What predicts a repeat: compare the 90-day repeat rate by first product or category, acquisition channel, first-order discount size, first-order value, and delivery experience if available (late or damaged). Show counts and percentage-point differences, and flag small groups.
6. Actions to test: for example a reminder timed just before the median reorder interval, a second-order incentive only for segments with low repeat rates, pushing first products that lead to repeats, a subscription offer for replenishable lines. Each with the metric that would show it worked and a holdout group.
7. Give runnable code (SQL or pandas) that computes the windowed repeat rates and time to second order.
</task>

<constraints>
- Always use a fixed window with eligible customers only; never report an unwindowed repeat rate as a trend.
- Differences between groups are associations; a first product with high repeat rates may attract a different kind of customer. Say so.
- Do not invent figures. With only a summary or sample, give the method and code and label any figures as illustrative.
- For products bought rarely by nature (furniture, mattresses), say that repeat rate is the wrong primary metric and suggest referral or cross-category purchase instead.
</constraints>

<output_format>
## Headline
The 90-day repeat rate, its trend, and the strongest predictor.

## Data checks
Bullets.

## Repeat rate
Table: First-order month | Eligible customers | Repeat in 30 days | 90 days | 180 days.

## Time to second order
Median plus a distribution table.

## Reorder intervals
Median interval and the order 2 to 3 to 4 continuation.

## What predicts a repeat
Table: Factor | Group | Customers | 90-day repeat rate | Difference.

## Actions to test
Numbered, each with metric and holdout.

## Code
One code block.
</output_format>

可替换内容

  • {{order_data}}
  • {{product_context}}

分类与标签

电商运营、客户复购、订单分析、留存运营、复购周期、运营实验

前往 AI 图片创作