Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Marketing Analytics › Troubleshooting

Retargeting and Brand Search Take All the Credit

Why retargeting and brand search appear to drive most sales: last-click models, audiences of near-buyers, demand made elsewhere and no holdout test.
Troubleshootingintermediate8 min readUpdated

First, compare the share of credit with the share of new customers

Credit concentration is a pattern, so read it against something that does not depend on attribution. Take a closed period and list, for retargeting and branded search, their share of attributed conversions and their share of the customers who were buying for the first time. Retargeting by definition reaches people who already visited, and branded searchers by definition already know the name. If both together claim most of the conversions but reach almost no one who was new to the business when the period began, they are collecting demand rather than creating it. The first-time buyers have to come from the order system or CRM rather than from analytics' new users, as the guide to new and returning customers explains. The hierarchy of evidence explains why a measured lift outranks an attributed figure and how much weight each can bear.

Then look for any holdout or lift result that already exists. If one has been run on either tactic, its measured lift is the best evidence you have and should sit beside the attributed figure in the report. If none exists, the attributed figure is an assumption. This page is about credit landing in too few places; if the problem is that the channels together claim more sales than exist, the channel conversions exceed sales page covers that over-count. Both sit in the analytics troubleshooting hub.

Diagnosis table

Diagnosis

SymptomCheckLikely causeFix
Final-touch tactics win in the default report; earlier channels look weakThe model each report uses, and a comparison across modelsLast-click attribution giving the final touch all the creditRead a multi-touch view beside last click; judge on tests
Retargeting reports very high returns from a small, frequently reached audienceAudience definitions, recency and frequency; share of audience already in checkoutAudiences full of people already about to buyExclude near-buyers and recent purchasers; cap frequency
Branded search conversions rise when other channels spend moreBranded search volume against spend in reach and prospecting channelsBrand search capturing demand other channels createdCredit creation through lagged and tested measures
Email and social look weak while direct and brand search look strongShare of email and social links that carry campaign tagsUntagged visits shifting credit to the last tagged touchTag every link you control
Nobody can say how many of the credited sales would have happened anywayWhether a holdout or lift test has ever been runNo holdout or incrementality testRun a holdout and set budget from its lift

Fix, verify and prevent each cause

Last-click attribution giving the final touch all the credit

Under a last-click rule, the touch nearest the purchase takes the whole conversion. For most journeys that touch is a retargeting ad or a branded search, so those two win every report built on the rule. The attribution models comparison shows how each model shares credit and where each one misleads.

Fix: put a multi-touch or data-driven view beside the last-click report, and compare how much each channel's credit changes between them; large swings mark channels whose value the default report hides.

Verify: prospecting and upper-funnel channels gain credit in the multi-touch view, and the team discusses both views rather than one.

Prevent: label every channel report with its model, and never set budgets from a single-model view without a test behind it.

Retargeting audiences full of people already about to buy

An audience of everyone who added to basket or visited checkout in the last few days is full of people who will come back and buy whatever the ads do. A platform reports a conversion when someone it reached converts within its window; Meta's help centre documents those attribution settings, including view-based windows that can credit an ad the buyer merely scrolled past. The prospecting versus remarketing guide sets out why Remarketing reports the best returns and adds the least.

Fix: exclude recent purchasers and people already in checkout, cap frequency, and size the audience to its pool rather than to its reported return.

Verify: reported returns fall somewhat, while total sales hold steady, which shows the excluded conversions were never caused by the ads.

Prevent: review audience definitions and frequency whenever retargeting budget rises faster than the audience grows.

Brand search capturing demand that other channels created

Nobody searches for a brand they have never heard of. A branded search is usually the last step of awareness built elsewhere: a video seen weeks earlier, a social post, a recommendation, a podcast. Attribution sees only the search, so the channels that created the demand look weak and brand search looks like the engine. The brand versus performance guide explains how this flatters capture, and the guide to choosing an attribution approach matches each measurement method to the decision it can support, putting budget questions with tests and mix models rather than with models of recorded paths.

Fix: track branded search volume against spend in reach and prospecting channels on a lag, and give those channels credit through tests or mix modelling rather than through last-touch reports.

Verify: when reach spend changes, branded search follows after a lag, confirming where the demand comes from.

Prevent: report branded search as an outcome of creation work as well as a channel of its own.

Untagged visits shifting credit to the last tagged touch

When newsletter links, organic social posts or partner links carry no campaign tags, the visits they produce arrive as direct or as an unknown referrer, and the next tagged touch, often a retargeting ad or branded search, takes the credit. Google's documentation on URL builders explains how tags carry source, medium and campaign into analytics, and the naming conventions and UTM guide sets the rules. If direct traffic is also unusually high, the direct traffic too high page finds where the evidence is being lost.

Fix: tag every link the business controls in email, social and partner placements, and audit the top landing pages for untagged entry.

Verify: email and social gain visits and assisted credit, while direct falls.

Prevent: make tagged links a launch requirement for every campaign, using a shared builder.

No holdout or incrementality test, so attributed equals assumed

Every cause above changes how credit is split; none of them answers the real question, which is how many sales retargeting and brand search add. Only a comparison with a group that did not see them can. The incrementality testing guide covers the designs, and the remarketing holdout test lays out a ready experiment for this case.

Fix: withhold retargeting from a random Holdout Group or run a controlled test on branded campaigns, and compare total sales between exposed and unexposed groups.

Verify: the measured Incrementality sits in the report beside the attributed figure, and budget follows the measured lift.

Prevent: repeat the test whenever audiences, offers or competitive conditions change, so the lift figure stays current.

A prevention routine for credit concentration

Credit drifts toward the last touch quietly, one budget review at a time, because each review rewards the channel with the best reported return. A routine that puts other evidence beside that return stops the drift. In every monthly channel report, show each channel's credit under two models and its share of new customers, so a tactic that wins on credit and adds few new buyers is visible at a glance. Keep a tagging audit in the monthly checks: the share of email, social and partner visits arriving with campaign tags, and the trend in direct traffic. Watch the size and frequency of retargeting audiences, and flag any month where retargeting spend grows faster than the pool it draws from. Track branded search volume against reach and prospecting spend on a lag. Twice a year, or whenever budgets shift materially, run a holdout on retargeting and a controlled test on branded search, and replace the attributed return with the tested one in planning. Finally, agree in advance that no channel's budget rises on attributed returns alone; requiring test evidence for increases protects the channels that create demand from being starved by the channels that collect it.

Illustrative example — not a real result

A beauty retailer tests its last-touch stars

A beauty retailer's last-click report credited retargeting with 38% of online sales at a return of 11 times spend, and branded search with 27% at 15 times. Prospecting video and social shared 9% at under 2 times, and the board proposed cutting them. Only 6% of retargeting's credited buyers were new customers. The team found that newsletter links carried no tags, so email visits arrived as direct and credit passed to the next retargeting ad. They tagged email, excluded buyers from the last seven days from retargeting, and ran a four-week holdout on 20% of the retargeting audience. Sales in the holdout were only 9% lower than in the exposed group, giving an incremental return of about 2 times. Prospecting budget was kept, and retargeting was capped at the audience's pool.

For brand search specifically, the comparison of branded and non-branded search campaigns covers separate targets for the two and how to design a test of brand bidding.

Common mistakes

Setting budgets from a last-click report. Retargeting everyone who visited, including yesterday's buyers. Treating branded search as a source of demand rather than the end of it. Leaving email and social links untagged. Moving money toward the highest attributed return without a test.

Frequently asked questions

Does this mean retargeting and brand search are worthless?

No. Both can add real sales: reminding a hesitant buyer, protecting the brand's own name from competitors, or recovering a basket abandoned by accident. The problem is that their reported results include a large share of sales that would have happened anyway, so their true contribution is smaller than the report says. A holdout test tells you how much smaller, and that figure, not the attributed one, should set their budget.

Will switching to a data-driven attribution model fix it?

It spreads credit more widely across the touchpoints the property recorded, which usually moves some credit away from the last click. It cannot credit touches it never saw, such as unclicked ads or untagged visits, and it still describes correlation rather than cause. It is a better reporting view, not an answer to what retargeting or brand search add.

Should we stop bidding on our own brand name?

Not on the strength of attribution reports in either direction. Whether branded search ads add sales depends on whether competitors bid on your name and how often searchers click the organic result instead. That is a question for a controlled test on branded campaigns, read on total branded clicks and sales rather than on the paid campaign's own conversions.

Terms used on this page

Frameworks

Where this page is used

Also related

Sources

  1. About attribution settings — Meta Business Help Centre
  2. [GA4] URL builders: Collect campaign data with custom URLs — Google Analytics Help

Related