Your team is reviewing two creator partnerships for a brand with products sourced in China. One appears to bring customers who return; the other seems to produce one-time buyers. Before renewing one and replacing the other, check when those customers first purchased. A customer observed for ten days has not had the same opportunity to return as one observed for three months.
ONCE recommends a fixed-window customer comparison before using repeat purchase rates to commission new creative or change partners. A shared reporting date does not provide a shared amount of follow-up.
A new product signal, with an important limit
Friendbuy's company announcement distributed through Business Wire on 3 November 2026 introduced general availability of Friendbuy Influencer. It says combining the offering with Referral and Loyalty can connect creator-sourced customers with later purchasing activity. Some coverage appeared on 4 November; that is not the original announcement date.
ONCE checked the release and official product description on 5 November. This is a supplier's capability claim, not evidence that installing software increases retention. Shopify's current Customers reports documentation groups customer cohorts around first-order dates. The method below is ONCE's analytical proposal, not a promise about either tool's default report.
Set the clock at the customer, not the campaign
Two October campaigns can acquire customers on different days. Exporting their reports together at month end does not equalise the time since those purchases. A partnership that started earlier may accumulate more repeat buyers simply because its customers have been observable for longer.
Choose an outcome and a window appropriate to the product. Reordering a consumable differs from buying an accessory or a second durable product. The absence of an immediate second lamp purchase does not, by itself, establish dissatisfaction. Keep these types of return purchase distinct rather than applying a universal industry interval.
The ONCE fixed-window check
| Input | Specify consistently | Why it matters |
|---|---|---|
| First-purchase origin | Each new customer's first paid order under a stated rule | A creator's publication date is not every customer's start date. |
| Data cutoff | The latest date covered, including known delays or gaps | A missing imported order is not evidence of no return. |
| Observation window D | The same elapsed interval from each customer's origin | Ten-day and ninety-day accumulated outcomes are different measures. |
| Fixed denominator | All qualifying new customers who have completed D | Do not admit early repeat buyers before admitting their equally young non-repeat peers. |
Define the rate as customers making a qualifying repeat purchase within D divided by the corresponding new customers fully observed for D. A customer who buys several times counts once in this customer rate; order frequency and revenue are separate measures. Agree the timestamps, timezone and qualifying-order rules before comparison.
Keep customers who have not completed D in a pending group, whether or not they have already returned. Their activity is still useful and should be reported separately. If nobody has completed the window, label the result unavailable for comparison, rather than zero.
A fictional ranking problem
Suppose A and B each contribute 100 identifiable new customers, with the same purchase age within each group. A has ninety days of follow-up and twenty repeat buyers. B has ten days and five repeat buyers. The apparent 20% versus 5% compares different windows.
Now suppose A had three repeat buyers by its tenth day. The common ten-day description becomes 3% versus 5%. All numbers are illustrative, not a benchmark, client result or forecast. The example invalidates a premature ranking; it does not establish that B is better or predict its ninety-day outcome.
Product mix, discounts, seasonality, market and subsequent customer care may differ. Equal follow-up is a necessary preparation for this comparison, not proof that a creator caused the observed difference.
Preserve the acquired group and append what happens
Record the source rule, creator, campaign, first-purchase date, initial product and new-customer identification status. An existing buyer using a creator's code for the first time is not automatically new to the brand. Keep unmatched identities visible as unknown, and apply one consistent rule to overlapping sources.
Preserve the original group and append later orders and refund status. Do not quietly remove customers because they refunded or have not returned, thereby improving the rate retrospectively. State how refunds affect order eligibility, revenue and each metric, and retain changes. Production teams can usually work from grouped summaries rather than personal contact details.
If the connection between records is incomplete, fix it or report the uncertainty before attaching a “poor audience” label to a creator or script. A tracking rule identifies an association; it does not demonstrate that the order would disappear without the partnership.
Make the next content decision conditional
For pending groups, review actual first-order issues and set a later review date. Do not commission retention videos merely to make an immature report look complete.
For mature groups with different outcomes, compare the first products, offers, markets and subsequent contact before diagnosing creative. If actual questions concern setup, maintenance or accessory compatibility, develop a content hypothesis around those questions and test it. A low rate alone does not identify a missing video.
For strong repeat behaviour, also review refunds, discounting and contribution. More orders do not automatically fund a higher creator fee, and product experience or post-purchase service may share responsibility. Specify production, publication and business objectives separately when renewing; do not promise a partner every dollar of uncertain future customer value.
A useful asset for an ongoing production partnership
For ONCE, the reusable asset is a record of source rules, observation windows, product conditions, customer questions and decisions. Connecting those records with content can support a more informed managed service. Whether it improves results must be established from real work.
At your next review, add four fields: first-purchase origin, common window, customers fully observed and customers pending. Then discuss whether the next job is new creative, clearer onboarding or further observation. Use ONCE's production solutions to define a specific content need rather than commissioning more footage from a premature creator ranking.
Sources and limits
Friendbuy's company release, “Friendbuy Launches Influencer and Affiliate Marketing Software, Bringing Referral, Loyalty, and Creator Programs Together on One Platform,” 3 November 2026, read in full through NEWSnet's Business Wire distribution; its official product page; and Shopify Help Center, “Customers reports,” current undated documentation. Checked 5 November. The fixed-window method and arithmetic are ONCE's proposals and examples. No private customer data, controlled test, retention lift or profit outcome is claimed. The original cover is conceptual.
