
A promotion can end with high sales and leave important questions unanswered. Did it bring in new customers, or did people who were already planning to buy make their purchases? Which group needed more attention? Were there returns that changed theoutcome? Total revenue alone cannot distinguish between these situations.
Analyzing a campaign by customer groups, also called cohorts, helps you understand what happened. Each group shares a defined characteristic, such as having made their first purchase during the promotion. The comparison is useful when ituses clear criteria and acknowledges what cannot yet be proven.
Choose groups that answer a specific question
Start with a simple distinction: new customers during the campaign and customers who had purchased before. If your question relates to channels, you can distinguish groups based on the recorded source, while keeping a category forunknown cases.
Avoid splitting the data so much that each group ends up with too few cases. Also, do not change the definition after seeing the results to favor a conclusion. Note how you will form the cohorts and what question you expect to answer beforepreparing the report.
Set the time period and keep track of the campaign conditions
Record the dates, included products, discount terms, and channels where the promotion was active. If anything changed during the campaign, keep a record of it: changing an offer or running out of stock can affect how you interpret theresults.
To review what happens afterward, use comparable windows starting from each customer's purchase. Someone who bought on the first day had more time to return than someone who came in at the end. A recent group cannot yet be assessed from the sameperspective as an older one.
Look at the full outcome, not just revenue
Review confirmed transactions, revenue net of discounts and returns, and any costs you can reliably assign. Include operational indicators such as pending inquiries, delivery issues, or support requests.
These data explain part of the post-sale experience. A group that generates more revenue may also require more work to deliver on what was promised. Keep estimates separate from verified figures so you do not present an approximatecalculation as a definitive result.
Interpret differences without inventing causation
If customers from one channel bought more, consider what other factors may have played a role: product choice, availability, timing, or prior experience with the brand. An observed difference alone does not prove that the channelcaused the outcome.
The same caution applies when comparing results with a week without a promotion. Demand can change due to seasonality, the calendar, or other factors. Present the analysis as evidence to inform decisions and new tests, noting its limitations whenthere is no sufficiently comparable basis for comparison.
End with a decision and a follow-up review
Decide what to keep, what to fix, and what you need to observe for longer. For example, you could clarify the offer, improve source tracking, or look into an issue that occurred repeatedly in one group.
Assign an owner and a date to each action. If you plan to study repeat purchases, decide when a reasonable period for your business will have passed. Avoid declaring lasting success or a lack of retention as soon as the promotion ends: some resultsrequire follow-up.
Customer and activity history organized in Onix Board can support this analysis. The goal is to turn every campaign into verifiable learning for the next one, with decisions that connect marketing, sales, andcustomer service.