Repeat Purchase
Part of Retention for low-frequency purchases
Measuring recommendations when repeat purchase is rare
Separate recommendation intent, sharing, referred enquiries and completed sales when existing customers rarely repurchase.
Measure recommendations as observable introductions and the resulting new-buyer outcomes. Report repeat buying by the original customer separately. A durable-goods buyer may recommend a product long before needing another one. Survey intent, a shared link, an enquiry and a completed order each answer a different question.
Define the stages
Write a rule staff can apply. A new buyer may report a personal introduction; a visit may arrive through a customer-shared route. Unless a public review can be connected to a particular introduction, keep it in a separate category.
| Stage | Count | Limit |
|---|---|---|
| Stated willingness | Respondents who say they would recommend | Intention is not an introduction |
| Observable sharing | Shares through a measurable route | A share may never reach a buyer |
| Referred enquiry | New people who report or use a recommendation route | The source may be missing or reported inaccurately |
| Referred purchase | Introduced people who complete a qualifying order | Attribution alone does not show the recommendation caused the order |
Choose one unit for each line: people, shares, enquiries or orders. One person can share several times, while a new buyer can encounter several influences. Do not add the stage counts as though they were unique customers.
Defining the Stages of Recommendation Measurement
- Stated willingnessRespondents who say they would recommend
- Observable sharingShares through a measurable route (e.g. tagged URL)
- Referred enquiryNew people who report or use a recommendation route
- Referred purchaseIntroduced people who complete a qualifying order
Record attribution and its gaps
A tagged share URL can identify visits through that URL in Google Analytics when tracking works. Use consistent campaign labels, check the landing page and keep personal information out of URLs and campaign parameters.
Ask new buyers neutrally how they heard about the business, allowing a recommendation and an unknown answer. Record a named introduction only when it is supplied through an appropriate process; matching surnames or addresses is not evidence of a referral.
Tracked visits are a partial signal, not a count of all recommendations. A buyer may hear about a product in conversation and search later. Tags may be absent or lost, and a tagged visit does not prove that a genuine personal recommendation occurred.
Report tracked visits, buyer-reported introductions and unknown sources separately. Deduplicate buyers before reporting completed orders.
For a referred order, apply a stated rule for completion, refunds and new-buyer identity. Include discounts, incentives and relevant variable costs before calling the activity commercially valuable.
Attribution Signals vs. Actual Outcomes in Recommendations
- Tracked visits via tagged URL
- Partial signal – does not prove genuine personal recommendation
- Buyer-reported introduction
- Self-reported; may include unknown or inaccurate sources
- Deduplicated completed orders
- Only count unique new buyers after removing duplicates
Choose the denominator and window
For an invitation programme, start with everyone eligible for the invitation, including those who do nothing. Show how many were sent the invitation, shared, introduced an enquiry and led to a completed order. Set an observation period suited to the purchase decision; people whose period has not ended remain pending.
For a business-wide report, show verified introductions and resulting new buyers over a defined period. Do not divide them by survey respondents or clickers unless the label says that is the population. Keep product groups separate when their buying decisions differ substantially.
To assess whether a prompt adds introductions, assign eligible customers to the prompt or usual experience before sending it. Compare introductions, new-buyer orders and contribution per assigned customer over the same window. If outcomes are too sparse for a useful comparison, report descriptive counts and uncertainty.
Key Metrics for Recommendation Reporting
- Invitations sent
- All eligible customers in an invitation programme
- Shared via tracked link
- Measurable shares using consistent campaign labels
- Referred enquiries received
- New leads from recommendation routes
- Completed referred orders
- Orders meeting commercial criteria (including discounts and costs)
Respect the people introduced
Let customers choose whether and how to share information. If the business proposes to contact a referred friend itself, assess the recipient’s permission and the rules for that channel before using their details.
Publish each report with its introduction rule, attribution routes, deduplication rule, follow-up cut-off, costs and unknown-source count.
Best Practices for Respecting Introduced Individuals
- Allow customers to choose how and whether to share
- Assess recipient permission before contacting referred friends
- Publish reports with full methodological transparencyInclude introduction rule, attribution routes, deduplication logic, cut-off period, costs and unknown-source count
