Ad recall is the proportion of people who remember seeing a particular advertisement, or advertising for a particular brand, some time after they were exposed to it. It is a memory test. The metric exists because an impression only records that an advertisement was delivered to a screen; it says nothing about whether anything reached the person in front of it. Recall is the oldest and cheapest attempt to close that gap, and it remains the most frequently reported outcome in brand lift studies run by Google, Meta, Amazon and independent panel vendors.

The term covers a family of measures rather than one number. Unaided recall asks respondents to name, without prompting, the brands they remember seeing advertised in a category. Aided recall prompts with a brand name or a list of brands. Recognition goes further and shows the advertisement itself, asking only whether it looks familiar. Each step adds cues and raises the score.

How recall is measured

The classic research design is a survey at a fixed interval after exposure. Day-after recall (DAR), the method popularised by Burke Marketing Research for television, contacted viewers the day after a commercial aired and asked whether they remembered advertising in the product category, then probed for what they could describe. The score was the share of the programme audience who could play back something specific from the commercial.

Digital measurement replaced the telephone call with an intercept. Most platform studies split an eligible audience into an exposed group and a control group withheld from the campaign, then serve both groups a short questionnaire. On YouTube, the survey appears as a single question ahead of a video. Google's Brand Lift documentation lists ad recall alongside awareness, consideration and association, and reports the result as lifted users, cost per lifted user and absolute, headroom and relative lift, according to the Google Ads Help Center. The measurement window is 10 days.

If 30% of the exposed group and 22% of the control group say they remember seeing advertising for a brand, absolute ad recall lift is eight percentage points and relative lift is about 36%. The control figure is never zero, because people misremember, confuse brands and see the advertiser's other campaigns; a raw recall percentage without a control hides that baseline.

Amazon applies the same design. Amazon Brand Lift, extended to Japan in April 2025, lists ad recall as one of six objectives and surveys its opt-in Shopper Panel.

Meta's estimated ad recall lift

Meta took a different route: it predicts recall rather than asking about it for every campaign. Facebook announced a Brand Awareness objective in September 2015, according to Tech.co, and a Meta developer blog post described it as using "real-time proxy metrics" to "maximize ad recall lift". The reporting fields that came with it, estimated ad recallers and estimated ad recall rate, refer to people Facebook estimates would recall the advertisement "within two days".

The current metric, estimated ad recall lift (people), is defined as an estimate of the number of additional people who may remember seeing the advertisements, if asked, within two days, according to documentation reproduced by the analytics vendor Emplifi. Meta's own help pages could not be retrieved for this article, so that wording is sourced indirectly. Meta has not published the model. A December 2016 Tech.co article said it was "believed" to combine reach with time spent viewing. As of February 2024, Ad Recall Lift was available as a performance goal only under the Awareness objective, according to consultant Jon Loomer's catalogue of Meta's 21 performance goals.

A Google or Amazon recall figure is a survey result from a randomised study. Meta's everyday estimate is a modelled number produced by the seller of the media.

Origin and evolution

Recall research grew out of print. Daniel Starch, a psychologist who taught at Harvard from 1920 to 1926, founded Daniel Starch and Staff in 1923 and published Principles of Advertising the same year. His readership studies took respondents through a magazine they had read and asked which advertisements they had seen, a recognition approach later reported under the labels noted, associated and read most. Gallup & Robinson, founded in Princeton in 1948 by George Gallup and Claude Robinson, built its business on recall measures, which ask people to bring an advertisement to mind with fewer prompts.

Television made recall an industry standard. Burke's DAR was the predominant copy-testing measure through the 1950s and 1960s, read as proof that a commercial could "break through" and lodge a brand message in long-term memory, and its adoption by Procter & Gamble made it a research staple. The argument over whether recall or recognition is the better gauge was set out by Herbert Krugman in "Memory without recall, exposure without perception" in the Journal of Advertising Research in 1977, and revisited by Erik du Plessis in the same journal in April 1994. Du Plessis found "virtually no empirical evidence" behind many of the strongest claims on either side.

Recall then lost its place at the top of copy testing. Procter & Gamble reviewed 100 split-cable tests over 10 years and found no significant relationship between recall scores and sales, and Leonard Lodish of Wharton reached a similar conclusion from test-market data. In 1982 a consortium of 21 agencies issued the Positioning Advertising Copy Testing (PACT) principles, which held that single measures are generally inadequate. Persuasion scores replaced recall as the headline measure.

Digital revived it. Google brought Brand Lift into AdWords on March 11, 2013, and in October 2018 rebuilt it with IRI around continuously fielded surveys. TikTok launched its own brand lift studies in June 2021, with recall among the standard questions.

Why it matters

Recall is the metric upper-funnel video and connected television budgets are most often sold on, because it moves faster than awareness or purchase intent. Sellers lean on it heavily. Kargo said its self-serve Karlo tool lifted ad recall 21% and aided awareness 10% for an Ad Council campaign, without disclosing method or sample. Disney said its Pause+ Trivia format delivered brand recall at "10 times industry benchmarks" in beta campaigns, against an undefined benchmark. Amazon cited a BrightLine study claiming a 33% increase in unaided recall for interactive formats, and LG Ad Solutions and MediaScience reported 138% higher unaided recall for interactive connected TV (CTV) ads.

Controlled studies are rarer. FreeWheel and MediaScience tested 436 participants and found unaided recall of 33% for one streaming ad against 12% for one social ad, with recall falling from 41% under a light ad load to 26% under a heavy one. FreeWheel sells the streaming technology that came out ahead.

Limitations and disputes

The central dispute is whether remembering an advertisement predicts sales. The split-cable evidence of the 1970s and 1980s said it did not, and Horace Schwerin's argument that a claim can be well remembered yet irrelevant to buyers anticipated later critiques. A 2004 Admap study of 60 Unilever commercials by John Kastenholz, Chuck Young and Graham Kerr concluded that recall-based pre-testing risked producing dull, uninvolving advertising. Les Binet and Peter Field's 2013 analysis for the IPA, The Long and the Short of It, found that pre-tested campaigns were less likely to achieve long-term business effects, according to a 2019 Ipsos summary.

Recall also measures memory of an advertisement, not of the right brand. In a 2026 study of more than 1,600 respondents across four countries, Cathy Nguyen of the Ehrenberg-Bass Institute found that about one-fifth of viewers attributed a comparative Pepsi print advertisement to Coca-Cola, according to the institute.

Response bias is a further problem. On Device, a UK measurement firm, said on October 8, 2025 that across more than 2,000 brand lift studies run between 2019 and 2025, ad recall gave inaccurate results in eight cases out of ten when compared with passively measured exposure. Existing customers reported recall 3.1 times more often than non-customers at comparable exposure. Alan Bantick, On Device's head of research, said ad recall "is not fit for purpose as a proxy for exposure". On Device sells a competing method.

Self-reported surveys carry their own limits. Adobe's July 2026 survey of 1,002 US consumers found that just 17% felt confident recalling the last three ads they saw within 24 hours, a figure about perceived memory rather than tested memory.

Attention vendors claim to bridge the gap, with mixed evidence. A PwC review of Lumen's attention platform, published on April 27, 2023, found that the relationship between predicted attention and brand lift metrics such as ad recall was "more complicated" than for clicks and conversions. A 2024 Kantar analysis of 873 campaigns found channel attention did not directly correlate with brand contribution.

Not the same as

Brand awareness measures whether people know a brand exists, whether or not they saw any advertising. Recall is tied to exposure to a campaign.

Brand lift is the experimental method; ad recall is one of the questions it asks.

Salience is the propensity of a brand to come to mind in buying situations, measured across many cues. Unaided recall uses one cue and refers to advertising, not to purchase occasions.

Share of voice is a brand's portion of category advertising spend or impressions. It is an input that may produce recall, not a measure of memory.

Recent developments

Upwave opened a closed beta of Predictive Brand Lift on September 9, 2026, modelling survey outcomes without fielding surveys. Google Ads API v25.1 added 24 read-only lift metrics in August 2026. Meta's measurement framework, built on 307 studies across 54 advertisers, reported that 55% of advertisers get contradictory results across tools.

Timeline

  • 1923: Daniel Starch founds Daniel Starch and Staff and publishes Principles of Advertising
  • 1948: George Gallup and Claude Robinson found Gallup & Robinson in Princeton, New Jersey
  • 1950s-1960s: Burke's day-after recall becomes the dominant television copy-testing measure, adopted by Procter & Gamble
  • 1977: Herbert Krugman publishes "Memory without recall, exposure without perception" in the Journal of Advertising Research
  • 1982: 21 agencies publish the PACT copy-testing principles
  • April 30, 1994: Erik du Plessis reviews the recall versus recognition debate in the Journal of Advertising Research
  • May 31, 2004: Kastenholz, Young and Kerr publish their 60-commercial Unilever study in Admap
  • March 11, 2013: Google launches Brand Lift in AdWords
  • 2013: Les Binet and Peter Field publish The Long and the Short of It for the IPA
  • September 2015: Facebook announces the Brand Awareness objective, optimised for estimated ad recall lift
  • October 2018: Google rebuilds Brand Lift with IRI
  • June 2021: TikTok launches Brand Lift Studies
  • April 27, 2023: PwC review of Lumen's attention platform published
  • April 30, 2025: Amazon Brand Lift extended to Japan
  • October 8, 2025: On Device reports ad recall inaccurate in eight of ten cases across 2,000 studies
  • 2026: Ehrenberg-Bass study finds about one-fifth of viewers misattribute a comparative Pepsi ad to Coca-Cola
  • July 6, 2026: Adobe publishes its brand recall report
  • September 15, 2026: FreeWheel and MediaScience publish streaming versus social recall findings
  • October 1, 2026: Kargo launches Karlo, citing a 21% ad recall lift

Summary

Who: Advertisers and agencies commission recall measurement; Google, Meta, Amazon and TikTok report it on their own inventory, while research firms from Starch, Gallup & Robinson and Burke to Kantar, Ipsos and Cint have fielded it independently.

What: The share of people who remember seeing an advertisement or a brand's advertising after exposure, measured unaided, aided or by recognition, and usually reported as lift between exposed and control groups.

When: Print readership studies began in the 1920s, day-after recall dominated television testing in the 1950s and 1960s, and platform brand lift tools revived recall from 2013, with Meta's modelled estimate following in 2015.

Where: In post-exposure surveys served in video players, social feeds and opt-in panels, and in platform dashboards and reporting APIs.

Why: Delivery metrics cannot show whether advertising registered in memory. Recall is a cheap, fast signal of that, but its link to sales has been contested since the 1970s, and misattribution, response bias and seller-reported figures limit what a single score proves.