The online store has a product feed, measurement and an account in Sklik. The campaign is running. Yet it can contain products with no margin, people who have already purchased, and data that nobody properly trusts. Technically, everything is “switched on”. From a business perspective, it is a bit of a wild ride.
I like dynamic retargeting precisely because it can do the boring work for a person: it reminds someone of a specific product they viewed. That does not mean I turn it on for every online store and then admire the low PNO. First I need to know whether the data, economics and campaign boundaries are right.
First, the product, event and order must match
The principle remains the same as in the original 2016 article: the system must connect the product in the feed with the product a person saw on the website. However, the current Sklik implementation no longer rests on old screenshots of the administration interface. According to the current Sklik dynamic retargeting documentation, the product is passed through the ViewContent event and its ID must match the ID in the product feed exactly.
This is the first check. Not the banner colour. If the IDs do not match, the system has nothing to pair. If the order value is sent incorrectly, it is impossible to assess return on investment sensibly. And if a completed purchase is not recognised, you will keep showing the customer something they already have at home.
- Verify availability, price, URL and ID in the feed.
- In the browser, check the events for product detail, basket and purchase.
- Compare the product identifier on the website with the feed character by character.
- Test an order and the exclusion of purchasers.
- Only then release the budget.
Implementation through a tag manager makes sense if the data layer is stable. I also discuss how to keep measurement under control in the article about Google Tag Manager.
I do not divide audiences by attractive numbers, but by decisions
In the old setup, I worked with windows of 0–3, 4–7 and 8–14 days and a decreasing cost per click. That was a specific experiment, not a law of nature. The purchase cycle should determine the window length. For fast-moving consumer goods, interest may fade within a few days. For more expensive equipment, people take longer to decide.
I would start simply:
- One audience of product visitors with a reasonable membership duration.
- Exclusion of completed orders.
- Separating main categories only when they have a different margin or business significance.
- A frequency cap, so reminders do not turn into stalking.
Sklik lets you filter product groups by feed attributes. Its official recommendations for product groups also warn against overly detailed segmentation. That matches my experience with automation in general: I add complexity only when it enables a different decision.
PNO without margin can look great and still lie
In the original 2016 test, PNO was 2.27% after nine days without optimisation. I regard it as the historical result of one specific campaign, not a promise or today’s benchmark. Nine days may also fail to capture returned orders, seasonality or longer decision-making.
To manage a campaign, alongside PNO I need to know at least the margin, cancellations, stock availability and the share of orders that would probably have arrived without advertising. I also look at new and returning customers. Advertising can report revenue, but the company needs to know what remains after the campaign.
When it is better not to turn on dynamic retargeting
- The feed is out of date or lacks stable identifiers.
- The conversion has no value or is sent more than once.
- You do not know the acceptable order cost based on margin.
- The range has only a few items and manual advertising expresses the offer better.
- Consent and operational responsibility for marketing measurement have not been resolved.
In that situation, it is better not to launch the campaign for a while. Saved budget is also a result.
Checking after launch matters more than the ceremonial click
In the first few days, I check whether the right products are displayed, whether purchasers are truly excluded and whether revenue in the advertising system makes sense in broad terms compared with orders. Then I monitor performance by category, device and time since the visit. I make one change, record it, and allow enough data to accumulate.
Dynamic retargeting is therefore not a magic button. It is a well-scalable reminder that rests on boring foundations. And those usually determine whether a campaign helps the business or merely produces a pretty report.
For a broader framework of paid acquisition, continue with the article on managing advertising by business objective. If you first need to align measurement and decision-making logic, you can look at how collaboration works.