A Retargeting Strategy That Annoys Less and Sells More
Someone views a product and keeps seeing the same banner for weeks, even after buying it. That consumes both budget and patience. Retargeting becomes more useful when it addresses a specific obstacle, responds to a changed situation and has a clear stopping point.

The objective is to help someone who has shown relevant interest take the next step. This guide concerns paid re-engagement through advertising platforms. A more considered strategy can reduce unnecessary exposure, but increased sales must be demonstrated in the specific account rather than promised in advance.
01Separate a visit from meaningful intent
A general article reader, someone comparing prices and someone with an unfinished order represent different intentions. Use customer journey mapping to identify hesitation and missing information. A page view alone may be too weak a reason for a direct sales message.
| Visitor situation | Useful message | When to change the message |
|---|---|---|
| Read a general article | A practical guide or service explanation | After stronger interest or window expiry |
| Viewed a specific product | Availability, use and a relevant objection | After purchase or availability changes |
| Left an unfinished purchase | Delivery, returns or a route to completion | After the order is completed |
| Submitted an enquiry | An appropriate agreed next step | Hand ongoing contact to sales |
| Already purchased | Care or a relevant add-on in a separate flow | Exclude from the original acquisition offer |
Start with a few segments that genuinely need different next steps. Dozens of tiny lists can complicate management and restrict delivery. An audience must be eligible for the chosen platform; analytics visitor counts are not the same as the number of people available for personalised advertising.
02Verify consent and audience entry
Test acceptance, rejection and changed choices in the cookie consent interface. Seeing a tag in the page source is insufficient: verify the events and signals actually sent. Do not populate audiences with sensitive information or put email addresses in ordinary URLs. Use approved data flows appropriate to the purpose.
Google requires consent and consent signals for relevant personalisation and remarketing uses in the EEA. The Sklik retargeting tag also uses consent information. A technical setting alone does not establish compliance with every legal requirement; it must reflect the visitor’s actual choice.
03Choose a window that fits the decision
The membership window should not automatically use the longest duration available. A simple accessory has a different buying cycle from business software. Examine time to purchase, offer availability and follow-up activity. A short window can miss slower buyers; a long one can keep reaching people who solved the problem elsewhere.
If you separate recent and older interest, make the groups mutually exclusive where the platform allows. “Last seven days” and “last thirty days” overlap unless the more recent group is excluded from the older one. Treat the chosen durations as hypotheses to test, rather than universal recommendations.

04Change the journey after a purchase
A purchase or qualified enquiry should update the relevant audience. Check that the event represents actual completion, is not duplicated and is connected to the intended campaign exclusion. Allow for processing delays and incomplete recognition across devices. No setup can promise that every subsequent impression stops instantly.
Sklik’s retargeting documentation describes exclusions at campaign and ad-group level, with campaign exclusions taking priority. A relevant add-on offer may be useful after purchase, but belongs in a separate journey rather than an unchanged advertisement for the item already bought.
05Manage frequency for the actual campaign type
There is no universally correct impression count. The period, buying cycle, creative and other channels all matter. Google documents frequency capping for Display and Video campaigns, while the same guidance states it is not supported for Demand Gen. Do not assume the same controls exist across campaign types.
Sklik allows daily frequency settings at campaign or ad-group level. A limit in one platform does not create a shared cap across Google, Sklik and email. Review available reach and frequency reports alongside customer feedback. Replacing the banner does not solve excessive exposure if the same person continues receiving too many offers.
- Group of 50 people2
- Group of 40 people5
- Group of 10 people20
Fictional audience over one period: 50 × 2 + 40 × 5 + 10 × 20 = 500 impressions across 100 people. The average is 5, although 10 people saw the ad twenty times. Not client data.
The model’s average looks moderate while hiding a heavily exposed group. Where detailed reporting is available, examine the distribution, time period and overlap. Adding the reach of two platforms does not give a deduplicated audience because some people may appear in both.
06Give the ad a useful job
Identify the obstacle: use, delivery, compatibility or how the service works. Address it with a specific message and a matching landing page. A generic slogan does little for someone who needs a delivery date. Effective offer copy explains the next step without inventing urgency.
Do not automatically discount every abandoned basket. That can reduce margin on purchases that would have happened anyway. For dynamic ads, regularly verify product data, prices and stock availability. More precise targeting cannot rescue an offer that is no longer available.
07Check who is actually being reached
Calling a campaign “retargeting” does not establish its targeting. Google explains that optimised targeting can reach people outside manually selected segments. If the objective is to evaluate re-engagement alone, check the available restrictions and expansion settings for that campaign type. An audience signal is not necessarily a delivery boundary.
| Control | What to verify | Why it matters |
|---|---|---|
| Audience membership | Entry event, window and updates | An old visit may no longer indicate intent |
| Exclusions | Completed purchases and list overlaps | A changed situation needs a changed message |
| Frequency | Available controls and their scope | One campaign does not govern every channel |
| Audience expansion | Whether people outside the list are reached | Results may not be pure retargeting |
| Creative and destination | Matching offer, price and availability | The ad must match the actual website |
Coordinate paid re-engagement with sales and appropriately authorised email communication. Our email remarketing guide covers first-party database workflows separately. A shared journey should prevent someone who has already resolved the issue from continuing to receive the original prompt.
08Evaluate incremental value and commercial quality
Retargeting reaches people already familiar with the brand. High reported ROAS does not by itself establish how many purchases the ads caused. In management reporting, distinguish attributed conversions, cancellations, returns and contribution after associated costs. View-through reporting and different attribution windows can also change platform comparisons.
Where volume and implementation permit, use a controlled experiment with a comparable group not exposed to the campaign. Account for overlap and the experimental tools available in the account. Switching ads off for a week and comparing with the previous week is weak causal evidence when seasonality or other campaigns have also changed.
09Start with one manageable journey
Choose a segment with a known obstacle and a recognisable completion event. Document audience entry, membership duration, exclusions, message, budget and evaluation. Test the actual visitor journey, including rejected consent and a completed purchase, before launch. Make understandable changes so the subsequent result can be interpreted.
Send Listify your product or service, main audience and current re-engagement approach. We can review audience logic, message continuity and measurement, then propose an order for improvements. Customer personal data is not needed for the initial description.
10Frequently asked questions
How many times should someone see an ad?
There is no universal number. Consider the time period, buying cycle, message and other channels. Check the controls for that campaign type and monitor outcomes alongside signs of excessive repetition.
Should every visitor be retargeted?
Not automatically. Identify relevant intent, appropriate data use and a useful next step. A broad list without a meaningful offer can consume budget without proportional value.
Why does a buyer still see the ad?
The completion event or exclusion may be missing, updates may be delayed, or devices may not be connected. Check the full chain from completed order to campaign settings.
Does high ROAS prove effectiveness?
It reports attributed value relative to ad spend under the platform’s settings. It does not establish incremental sales or profit. Include commercial costs and, where feasible, test causal impact experimentally.
When should the strategy be reconsidered?
When the product, buying cycle, available data or sales process changes. Also reconsider it when higher spend primarily produces repeated exposure to the same small audience.