Quick Answer
Yes, retargeting still works in 2026, and it is often the cheapest conversion line in a Canadian ad account. What changed is the supply of people you are allowed to retarget. Chrome kept third party cookies, Google retired the Privacy Sandbox replacements, Safari and Firefox still block cross site cookies outright, and Quebec consent rules now gate the tag itself. Smaller pools, same mechanics.
By Cody Wise, founder of Wise Media. Published 17 November 2026. This article explains advertising and privacy mechanics for operators. It is not legal advice, and privacy obligations differ by province and by what your business collects. Confirm your own position with a Canadian privacy lawyer before relying on it.
Summary
- Retargeting works, but on a smaller base. Your addressable pool is a fraction of your traffic, and the gap is consent plus browser policy, not ad platform quality.
- The cookie deadline is gone. Google confirmed on 17 October 2025 that it is retiring almost all Privacy Sandbox technologies and keeping its current approach to third party cookie choice in Chrome. There is no migration date to plan around any more.
- Consent is now the binding constraint in Canada. In Quebec, profiling technology has to be off until a person turns it on. Outside Quebec, PIPEDA still requires meaningful consent for behavioural advertising.
- Audience size is the real failure mode. Most small Canadian accounts cannot fill a useful retargeting pool from organic traffic alone, so the fix is upstream traffic, not a new audience setting.
- Retargeting takes credit it did not earn. Last click reporting flatters it. The only honest read is a holdout or geo test.
- Google removed the shortcut. Similar audiences were deleted from all campaigns in August 2023. First party lists plus optimized targeting replaced them.

Table of contents
- Does retargeting still work in 2026?
- Why your retargeting pool is smaller than your traffic
- What Canadian consent rules require before you can retarget
- How to tell whether retargeting is working or just taking credit
- What retargeting should cost, and how to budget it
- Seven mistakes that make retargeting look dead
- The retargeting rebuild, in order
- FAQ
Does retargeting still work in 2026?
Retargeting still works. On most Canadian service and ecommerce accounts it returns the lowest cost per conversion of any campaign type, because it only speaks to people who already showed intent. What has collapsed is not the mechanic. It is the number of people the mechanic can reach, and the certainty with which you can count them.
If you are running paid traffic in Canada and retargeting has quietly stopped producing, the cause is almost always one of three things: the pool is too small to serve, consent is blocking the tag, or the campaign was never incremental in the first place. We rebuild exactly this layer inside our paid advertising service, and the diagnosis order below is the one we use.
What actually changed between 2024 and 2026
Three years of industry warnings about the end of cookies produced an outcome almost nobody planned for. The replacement failed, and the thing it was supposed to replace stayed.
On 17 October 2025, Google published an update stating it is retiring the Attribution Reporting API, Protected Audience, Topics, Private Aggregation, Related Website Sets, IP Protection, On-Device Personalization, Protected App Signals, SelectURL and the SDK Runtime, citing low levels of adoption. In the same post it said Chrome will maintain its current approach to offering users third party cookie choice. CHIPS, FedCM and Private State Tokens survive. You can read it at Google’s Privacy Sandbox blog.
For an advertiser that means something very specific. There is no deprecation date to migrate before. Chrome, which carries the large majority of Canadian desktop and Android traffic, still supports the third party cookies that classic display retargeting depends on. The doom scenario did not arrive on schedule, and the sanctioned replacement for it no longer exists either.
What did not change, and why it matters more
Safari and Firefox never waited for Google. Both block cross site tracking cookies by default and have for years, which means the slice of your Canadian traffic arriving on iPhone Safari was already invisible to cookie based retargeting before any of this started. Apple’s App Tracking Transparency rules compound it inside apps: Apple defines tracking as linking your app’s user or device data with data collected from other companies’ apps or websites for targeted advertising, requires permission through the ATT prompt before you may do it, and states that without permission the advertising identifier returns all zeros and you may not track. Apple also rules out working around a refusal with hashed emails, fingerprinting, or a second consent prompt. Its policy is published at Apple’s user privacy and data use page.
So the honest 2026 picture is not that retargeting died. It is that retargeting now reaches a consent-filtered, browser-filtered subset of your visitors, and that subset is both smaller and systematically different from your audience as a whole. Skewed toward Chrome, toward Android, toward desktop, and toward people who clicked accept.
| Channel | Still viable in 2026 | What it depends on |
|---|---|---|
| Google Display and YouTube remarketing | Yes | Chrome third party cookies plus your tag, consent signals, 30 day minimum pool |
| Meta website retargeting | Yes, with gaps on iOS | Pixel plus Conversions API, ATT status, consent |
| Customer Match and Meta customer lists | Yes, strongest option | First party email and phone data, lawful consent to use it for advertising |
| Search remarketing lists (RLSA) | Yes | Pool size, and people searching again |
| Third party data segments bought from brokers | Weakest | Broker data quality, and rising legal exposure |
| Similar or lookalike expansion from a seed list | Changed | Google deleted similar audiences in August 2023. Optimized targeting replaced it |
That last row is the one most accounts never adjusted to. Google’s own documentation states that new similar audiences segments stopped being generated in May 2023 and that from August 2023 similar audiences segments were removed from all ad groups and campaigns, with optimized targeting, audience expansion and Smart Bidding presented as the replacements. The source is Google’s help documentation on durable audience strategies. If your structure was built on a seed list plus a similar audience, half of it has been silently gone for three years.
Why your retargeting pool is smaller than your traffic
A retargeting list is not a copy of your analytics traffic. Every visitor passes through a series of filters before they land in an addressable audience, and each filter removes a share. Nothing is broken when this happens. It is the design of the modern web.
Where the signal leaks
| Filter | What it removes | What you can do about it |
|---|---|---|
| Consent banner refusal or non-interaction | Anyone who declined or ignored the banner, where your banner gates marketing tags | Fix banner design and placement. Implement consent mode so modelled conversions still arrive |
| Safari and Firefox cross site cookie blocking | Cookie based display and social retargeting for those browsers | Shift weight to Customer Match, server side tagging, email and search remarketing |
| iOS App Tracking Transparency refusals | In-app identifier based retargeting for users who said no | Conversions API, first party lists, and accept the measurement gap |
| Ad blockers and privacy extensions | Client side tag fires entirely | Server side tagging recovers part of it |
| Membership duration expiry | Visitors older than your window, commonly 30 or 90 days | Lengthen the window for considered purchases, shorten for impulse |
| Platform minimum audience thresholds | The whole campaign, which simply will not serve | Combine pages into one list rather than splitting into ten micro audiences |
The compounding effect is what surprises people. A Canadian site with 4,000 monthly visitors does not get a 4,000 person retargeting pool. After consent refusals, browser blocking, blockers and the membership window, it may get a few hundred addressable users spread across two platforms, which is often below the floor where a display campaign will serve at all. The campaign then spends nothing, reports nothing, and gets labelled dead.
The diagnostic question that settles it in ten minutes
Open the audience manager in Google Ads and Meta Ads Manager and read the size of each list you are actually targeting. Not your traffic, not your sessions. The list size. Then compare it against your monthly sessions for the same window.
- List size under roughly 1,000 and falling: this is a traffic problem, not a retargeting problem. Retargeting cannot manufacture demand it never saw.
- List size a small single digit percentage of sessions: this is a tagging or consent problem. Something is stopping the tag from firing for most visitors.
- List size healthy, performance poor: this is a creative, frequency or offer problem, and the rest of this article is where to look.
If the answer is the second one, the recovery path is server side tagging and a properly wired consent layer rather than a new audience. We covered the build end to end in our guide to server side tracking for Canadian businesses, and the consent signalling piece in Google consent mode in Canada.
What Canadian consent rules require before you can retarget
Canada has no single national cookie law. It has a federal privacy statute that applies to commercial activity, a stricter provincial regime in Quebec, and three provinces with substantially similar private sector laws. For retargeting, the two that decide your build are PIPEDA and Quebec’s Law 25.
Outside Quebec: meaningful consent under PIPEDA
PIPEDA requires knowledge and consent for the collection, use and disclosure of personal information in the course of commercial activity, and the Office of the Privacy Commissioner has long treated behavioural advertising as an activity that requires consent people can actually understand and withdraw. In practice, for a retargeting setup, that means your privacy policy names behavioural advertising, your banner is not a dark pattern, and withdrawal actually stops the tag. If your privacy policy predates your pixel, start with whether you need a privacy policy in Canada.
Quebec: profiling technology off until switched on
Quebec is the stricter case and the one most Canadian sites get wrong. Law 25 added section 8.1 to the Private Sector Act, covering technology that allows a person to be identified, located or profiled. Businesses must inform people of the technology and of the means to activate those functions. McCarthy Tetrault’s TechLex analysis, published 11 October 2023 by Michael Scherman and Wendes Keung, notes that the statute does not use the words deactivated by default, but that the drafting change from deactivate to activate implies profiling functions must start off, and recommends consent before any cookie that is not strictly necessary. That reading is a law firm’s interpretation of a provision that has not been fully tested, and we flag it as such rather than as settled law. Most of the amendments came into force on 22 September 2023.
Advertising cookies tailor what a person sees based on browsing history, which is profiling on any ordinary reading. If you serve Quebec and your retargeting tag fires before consent, that is the exposure. Our breakdown of Law 25 cookie consent requirements goes through the banner mechanics.
The build that satisfies both without destroying the data
- Block marketing and analytics tags until consent, by category, not as one all or nothing switch.
- Run consent mode so the platforms receive a signal even on refusal and can model what they cannot observe.
- Put the banner where people read it, with accept and decline given equal visual weight. A banner engineered to trick people is both a legal risk and a data quality risk, because forced accepts produce audiences that do not convert.
- Name behavioural advertising in the privacy policy in plain language, and make withdrawal real.
- Keep a dated record of your consent configuration. If anyone ever asks what your site did on a given date, that record is the answer.
There is a commercial upside hiding in this. A consent layer built properly produces a smaller but cleaner audience, and cleaner audiences bid better. Accounts that force consent tend to carry pools full of people who never wanted the ad.
How to tell whether retargeting is working or just taking credit
Retargeting is the most flattered campaign type in digital advertising. It shows ads to people who already decided to buy, then claims the sale. A reported return of eight to one can mask a true incremental return close to zero. The fix is not a better attribution model. It is a test.

Run a holdout, not a report
An incrementality test answers one question: would these conversions have happened anyway? There are two practical versions for a small Canadian account.
- Platform conversion lift, where available. The platform withholds ads from a randomised share of your audience and compares outcomes. Cleanest method, but minimum spend and conversion volume requirements often put it out of reach for smaller accounts.
- Geo holdout. Split comparable markets, for example run retargeting in Calgary and Edmonton and pause it in Winnipeg and Halifax for four to six weeks, then compare total conversions per market against the prior period. Crude, but it uses data you already have and it cannot be gamed by attribution settings.
A third option that costs nothing: pause retargeting entirely for two weeks and watch total account conversions, not campaign conversions. If total volume does not move, the campaign was reassigning credit rather than creating sales. Do this in a stable period, not during a promotion or a seasonal swing.
Why the platforms will never agree
Google, Meta and GA4 each count a retargeting conversion using different attribution windows, different view-through rules and different modelling. On retargeting specifically the disagreement is at its widest, because view-through and engaged-view conversions are a large share of the total. Do not reconcile the three. Pick one source of truth for decisions, usually your own CRM or order data, and treat platform numbers as directional. We went through the mechanics in why Google Ads, Meta and GA4 never agree on conversions.
What retargeting should cost, and how to budget it
Retargeting is cheap per impression and expensive per mistake. The common error is treating it as a budget line rather than a function of audience size. You cannot spend a thousand dollars a month against four hundred people without burning them.
Size the spend to the pool, not to the ambition
Work backwards from frequency. Decide how many times you are willing to show one person your ad in a week, usually three to five for a considered purchase and fewer for a low value one. Multiply by pool size to get weekly impressions, then apply your platform CPM to get a defensible budget. If that number comes out at a few hundred dollars CAD a month, that is the correct answer, and adding more would simply raise frequency until people resent the brand.
- Allocation that usually holds: most of the budget on acquisition, a minority on retargeting. Retargeting is a conversion tax on traffic you already paid for, so it scales only when acquisition scales.
- Window length: 7 to 14 days for impulse and ecommerce, 30 to 90 days for services with a long consideration cycle, longer again for high ticket construction or professional work.
- Segment by intent, not by page count: three segments is usually the ceiling for a Canadian small business. Viewed a service page, started a form or added to cart, and existing customers. Ten micro segments starve each other.
- Frequency caps: set them. The default on most placements is higher than any sensible brand wants.
For how this sits inside a whole paid budget rather than on its own, see our breakdown of Google Ads budgets for Canadian small business.
Where the money is better spent instead
If your retargeting pool is small, the highest return move is almost never a retargeting change. It is one of these:
- Email and SMS capture. A first party list is the one audience nobody can deprecate. It survives browser policy, consent resets and platform changes, and it loads straight into Customer Match.
- Landing page conversion rate. Lifting a page from two percent to three percent does more than any audience setting, and it compounds across every channel. Start with why Google Ads clicks are not converting.
- Lead capture mechanics. Whether the page asks for a booking or a form changes the pool more than the pixel does. See booking calendar versus contact form.
- Organic and AI search presence. Retargeting needs a top of funnel to retarget. If yours is thin, the SEO map and the order to run disciplines in is the better next project.
Seven mistakes that make retargeting look dead
- Retargeting everyone who touched the site. A visitor who bounced from a blog post in nine seconds is not in market. Exclude short sessions and single page bounces, and your cost per conversion usually drops without any other change.
- Never excluding converters. Showing the same offer to someone who already bought it wastes budget and makes the brand look careless. Build a converter exclusion before anything else.
- One creative, forever. Retargeting runs against a fixed pool, so the same person sees the same ad repeatedly. Three to four creatives on rotation is the minimum, refreshed roughly monthly.
- Repeating the acquisition message. The person already saw that. A retargeting ad should answer the objection that stopped them: price, timeline, trust, or scope.
- Splitting into ten micro audiences. Each one falls below the serving threshold and none of them spend. Consolidate.
- Counting view-through conversions as real. They inflate reported return more than any other setting. Decide whether you believe them before you decide the campaign works.
- Still running on a similar audience structure. Google removed those segments from all campaigns in August 2023. If your account was built before that and nobody rebuilt the targeting, part of it has been inert for years.
Retargeting in 2026: pros and cons
| Strengths | Weaknesses |
|---|---|
| Lowest cost per conversion in most accounts | Takes credit for conversions that would have happened anyway |
| Short feedback loop, results inside two weeks | Cannot scale beyond the size of your traffic |
| Excellent for recovering abandoned carts and stalled quotes | Fatigues a small pool quickly and can damage brand perception |
| Customer Match lists are durable against browser and platform change | Requires lawful consent to use customer data for advertising |
| Cookie based display still functions in Chrome with no deadline | Invisible on Safari and Firefox, and gated on iOS apps by ATT |
| Cheap to test, cheap to pause | Measurement is genuinely hard without a holdout |
The retargeting rebuild, in order
Run these in sequence. Each step tells you whether the next one is worth doing.
- Read your actual audience sizes in Google Ads and Meta. Write the numbers down next to your monthly sessions.
- Audit the consent layer. Confirm marketing tags are blocked before consent, that decline works, and that consent mode is sending a signal either way.
- Confirm the tag fires on the pages that matter, in a real browser, with the consent banner accepted and declined. Test both states.
- Collapse your segments to three or fewer, each above the platform serving threshold.
- Build the exclusions: existing customers, recent converters, bounced sessions under ten seconds.
- Load first party lists into Customer Match and Meta customer lists. Confirm you have consent covering advertising use.
- Set frequency caps and window lengths deliberately, matched to your sales cycle.
- Write objection-led creative. Three to four variants, each answering a different reason people stall.
- Run a holdout for four to six weeks. Geo split if platform lift is unavailable.
- Decide with the holdout number, not the platform number. Scale, hold, or cut.
If steps one through three reveal the pool is tiny because traffic is tiny, stop. The project is demand generation, not retargeting, and the lead quality problem is usually sitting upstream of it.
Frequently asked questions
Is retargeting dead now that third party cookies are going away?
No, and the premise changed. Google announced on 17 October 2025 that it is retiring most Privacy Sandbox technologies and keeping its current approach to third party cookie choice in Chrome. Cookie based retargeting continues to work in Chrome. Safari and Firefox still block it, which they already did.
How big does a retargeting audience need to be?
Each platform publishes its own minimum and they change, so check the current threshold inside your account rather than relying on a number from an article. As a working rule, if a list is in the low hundreds it will serve poorly or not at all, and consolidating segments is more useful than adding budget.
Do I need a cookie consent banner in Canada to run retargeting?
If you serve Quebec, treat it as required: Law 25 section 8.1 covers technology that can identify, locate or profile a person, and the prevailing legal reading is that those functions must start off. Outside Quebec, PIPEDA requires meaningful consent for behavioural advertising, which in practice means a real notice and a working withdrawal. This is general information, not legal advice.
Why did my retargeting conversions drop without any campaign change?
Check four things in order: a new or changed consent banner now blocking the tag, a site migration that dropped the tag, membership duration expiring faster than new visitors arrive, and an audience that fell below the serving threshold. In our experience a consent banner change is the most common single cause.
What replaced Google’s similar audiences?
Google’s documentation names optimized targeting for Demand Gen, Display and Video action campaigns, audience expansion for video reach campaigns, and Smart Bidding for Search and Shopping, all fed by your first party data. Similar audiences segments were removed from all ad groups and campaigns in August 2023.
Is server side tracking worth it just for retargeting?
Only once the account is big enough that the recovered signal pays for the setup and monthly hosting. For a small account the same money spent on landing page conversion rate or email capture usually returns more. The decision framework is in our server side tracking buyer’s guide.
Should I buy third party audience data instead?
It is the weakest option on the list and carries the most legal exposure, because you inherit consent questions about data you did not collect. Building a first party list is slower and worth more.
The bottom line
Retargeting in 2026 is a smaller, cleaner, better measured version of what it was. The cookie cliff never arrived, the replacement for cookies was cancelled, and the real constraints turned out to be consent, browser defaults and audience size. Accounts that treat retargeting as a conversion tax on traffic they already earned do well. Accounts that treat it as a growth channel keep being disappointed, because it was never one.
Audit the pool, fix the consent layer, collapse the segments, run a holdout. Four steps, and most of the mystery disappears.
Want this rebuilt properly?
Wise Media builds the measurement and retargeting layer for Canadian founders and agencies, from the consent configuration through to the holdout test that tells you whether to scale. If your retargeting has quietly stopped producing and you want a straight answer about why, start with our intake form and we will tell you what we would fix first.



