Retargeting a cash-on-delivery store: who belongs in each audience?
Retargeting does not mean chasing every visitor with the same purchase advertisement. Someone seeking dimensions differs from a customer awaiting confirmation; a delivered buyer differs from someone who cancelled because the product was no longer needed. Start with each group’s remaining question, then choose a suitable message and
Retargeting does not mean chasing every visitor with the same purchase advertisement. Someone seeking dimensions differs from a customer awaiting confirmation; a delivered buyer differs from someone who cancelled because the product was no longer needed. Start with each group’s remaining question, then choose a suitable message and boundaries that prevent unnecessary pressure or expenditure. Installing a pixel guarantees neither a usable audience nor profitable results.
Segment by real stage
Begin with four understandable situations: interest without an order, an active order, completed delivery, and cancellation or refusal. This is a commercial thinking map, not a promise that an advertising platform automatically knows every status. Some events describe browser behavior, while COD stages live in order management and may arrive later. Unless suitable information reaches the audience tool through an available permitted mechanism, do not assume an exclusion works automatically.
Watching a video does not establish firm purchase intent either. That viewer might need a demonstration, while someone who already ordered needs service follow-up. Separating these needs makes messages more relevant.
Answer a particular unresolved question
| Situation | Possible useful message | Avoid |
|---|---|---|
| Interest without an order | Dimensions, use or clear delivery costs | Claiming to know why the person left |
| Order being processed | Appropriate confirmation and service | Suggesting another order is required |
| Completed delivery | Usage guidance or a relevant complement | Repeating acquisition offers without reason |
| Cancellation | Understanding the reason first | Pressure or insulting labels |
Avoid intrusive statements suggesting personal surveillance, and do not infer health, beliefs or sensitive circumstances. You can explain product suitability without turning private information into advertising language.
Make exclusions realistic
Excluding people who completed the intended action can be useful for the same acquisition campaign when your account has suitable information and functionality. It does not mean removing every buyer from all future marketing forever. A relevant complementary product may make sense later. Record the exclusion reason and duration, then review data freshness. Do not use purchased contact lists or customer information without an appropriate basis and usage rights.
Matjri helps organize order statuses and distinguish confirmation from delivery. That does not establish automatic synchronization of every exclusion list into Meta or TikTok. The current native merchant measurement integration supports Meta. Another advertising platform requires an actually available, tested solution rather than simply swapping the pixel identifier.
A fictional audience and cost example
Imagine 200 product visits and 20 known orders:12 delivered,4 being processed and 4 cancelled. Do not assume you have 200 people available for retargeting. Visits are not necessarily distinct usable audience members, and signals may be missing. Subtracting order states on a worksheet does not mean the corresponding people are technically excluded in the advertising tool.
Suppose an illustrative 60 TND experiment explaining dimensions produces 3 deliveries linked to the message in your records. Advertising costs 20 TND per delivery. If contribution before advertising is 18 TND, the result does not cover the advertisement even before other costs. Calling an audience warm does not make every acquisition cost acceptable.
Run a reviewable experiment
- Write the audience stage and unanswered question.
- Identify recent information that can appropriately be used.
- Prepare one message and a consistent destination.
- Check exclusions actually available in the account rather than assuming they operate.
- Set spending and review limits, then examine orders, complaints and repeated exposure where available.
A very small audience can produce unstable results or see the same advertisement too frequently. Greater pressure is not the solution. Review data volume, objective and platform eligibility. Improving the destination, publishing useful organic content or providing appropriate service may be preferable to creating a separate campaign.
Keep customer service separate from sales pressure
A customer waiting for a delayed parcel needs a solution before another discount advertisement. Use order records so the person does not repeat the same issue to every employee. Respect requests to stop marketing communication within your procedures. Cancellation is not proof of bad character; an unclear offer or unsuitable timing may explain it.
Your next step is sorting ten previous orders by outcome and writing one useful question for each situation, without uploading a list before checking applicable conditions. If your records need structure, begin with product and order management in Matjri, then design retargeting whose limitations you understand rather than relying on an audience label.
Frequently asked questions
Should everyone who ordered be excluded, even after cancelling?
That depends on campaign purpose, cancellation reason and data age. Distinguish an ongoing order from a request not to be contacted. Write a clear rule and verify the tool can implement it.
Is a smaller audience always cheaper?
No. Available people may be scarce, exposure repetitive and costs high. Compare commercial results with your risk limit instead of assuming a known audience is easy to convert.
Does connecting a pixel build every audience automatically?
No. Events, information, eligibility and configuration still matter. A status in your store does not automatically reach every marketing service unless an actual transmission mechanism exists and is appropriately used.
Sources and review
Editorial review: 15 October 2026. Numerical examples are illustrative learning exercises, not promised results or market prices.