Targeting That Got Blunter

What has actually happened

The tracking permission prompt arrived in the spring, and enough time has now passed to say what it did rather than what was predicted.

A large majority of people decline when asked, which was expected. What follows from that has been more gradual.

Audiences built from site visitors and app users are considerably smaller than they were.

Reported results under-count, sometimes substantially, because a proportion of conversions can no longer be observed.

And targeting itself has become less precise, because the signals that made it precise are thinner.

The three practical effects

• Smaller audiences built from your own visitors.
• Shorter attribution windows, so later conversions go uncounted.
• Delayed and modelled reporting rather than direct counting.
• Weaker lookalikes, built from thinner source data.
• Fewer events you can optimise toward.

The second is the one that misleads people most, because a campaign can look worse than it is when conversions happening a week later are simply not attributed.

The reversal nobody expected

The genuinely interesting development, and it is counterintuitive.

For years the received approach was to narrow the audience as far as possible, on the reasoning that precision reduced waste.

With thinner signals, narrow targeting now frequently performs worse than broad targeting, because you are constraining the system using data that is less reliable than it was, and preventing it from finding people it might otherwise have found.

Broad targeting with a strong message, letting the system optimise on outcomes, is now outperforming carefully layered interest targeting in a great many small accounts.

That is a real reversal of a decade of practice, and it is worth testing rather than assuming, since it does not hold everywhere.

Why broad can beat narrow now

The reasoning behind it, briefly.

Advertising systems find people by observing who responds and looking for more of them, which works better with more room to search.

Narrow targeting was useful when the interest and behaviour data was rich enough to be a genuine shortcut.

As that data thins, the constraint becomes a handicap: you are excluding people on the basis of categories that are less accurate than they were, using your own assumptions about who your customer is.

Those assumptions are frequently wrong in ways the system would have discovered.

The practical translation is to define the audience by location and any genuine hard constraint, and let the optimisation do the rest.

A worked example

A small advertiser had run the same structure for two years: tight interest stacks, several layered conditions, and a modest daily budget.

Through the summer, cost per enquiry rose steadily and volume fell.

They tested one campaign with the targeting removed entirely apart from a location radius and an age range, optimised toward form submissions, with the same budget and a clearer message.

Cost per enquiry came in below the narrow campaign within about three weeks.

They kept both running for a further month before shifting most of the budget across.

The part they found hardest was accepting that two years of careful audience work had become a liability rather than an asset.

Reporting under-counts

Worth handling deliberately, because decisions made on incomplete numbers compound.

The platform's reported conversions are now a partial and partly modelled figure, so a campaign showing eleven conversions may have produced more.

Which means comparing this year's reported cost per acquisition against last year's is comparing two different measurements and concluding something false.

The remedy is a count on your side: enquiries received, calls taken, orders placed, and a simple question asking people how they found you.

Compare total enquiries against total spend across a month, which is cruder than platform attribution and considerably more honest.

Businesses that made that shift this year are the ones still able to make sensible decisions.

Rebuild what you can observe

The durable response rather than a workaround.

Information customers give you directly is unaffected by any of this: an email address, a phone enquiry, a purchase record, an answer to how did you hear about us.

That data can build audiences, can be uploaded for matching, and can be used to reach people without any intermediary.

It also improves rather than degrades over time, which is the opposite of what is happening to tracking-based audiences.

The businesses least affected by this year are the ones that were already collecting it, which was true before the change and is simply more visible now.

What to stop doing

Three habits that are now actively unhelpful.

Reacting to short-term reported fluctuations, which are noisier than they were and will send you in circles.

Making changes weekly, since attribution delays mean you are frequently judging a campaign on incomplete data.

And rebuilding narrow audiences repeatedly in the hope of recovering earlier performance, which spends effort on the part that has structurally changed.

Give campaigns longer, judge them on your own numbers, and change fewer things at once.

The counter-case

Broad targeting is not universally better and the advice has been overstated in places.

Where your addressable market is genuinely narrow, such as a specific trade, a small geography, or an expensive niche product, broad targeting wastes money on people who could never buy.

Larger accounts with substantial volume also retain enough signal for narrower approaches to work, since the degradation matters most where data was already thin.

And an account with a weak offer or a poor landing page will fail at any targeting setting, which is worth ruling out before restructuring anything.

Test it in your own account rather than adopting it because it is being said everywhere, which is the only reliable way to know.

What to do now

1. Test one broad campaign against your narrow structure.
2. Optimise toward the real outcome, not a proxy.
3. Give it three weeks before judging.
4. Count enquiries on your side.
5. Ask customers how they found you.
6. Compare monthly totals rather than platform attribution.
7. Collect direct data at every opportunity.

Step four is what makes every other decision possible, and it costs one question asked consistently.

The earlier version of this is covered in rebuilding a campaign around less data.

Frequently asked questions

West Coast Media Solutions Inc. provides web design, web development, hosting, digital marketing, and business consulting to organisations across Canada, drawing on more than twenty-five years in the field.