Paid Media: 109 findings, with sources

CMOs are prioritising brand while planning fewer traditional brand campaigns than ever

CMOs are moving money toward brand while moving away from the format brand money used to buy: for the first time, under 50% are planning traditional brand campaigns. The budget is being reallocated to creator and video formats rather than to campaigns in the old sense.

Shares of content that helps people escape uncertainty more than tripled during 2025

Shares of content that helps people escape uncertainty rose 214% between the first and second half of 2025, measured across 1.6 trillion public shares on Facebook, Instagram and Threads.

Comedy is the most shared content category on social, ahead of motivation

Comedy was the most shared content category of 2025 with 173 billion shares globally; motivation was third with 119 billion.

Two in five CMOs cannot measure the efficiency of next-gen video at all

43% of CMOs say they are struggling to measure the efficiency of next-gen video formats - the measurement gap that pushes budgets back toward whatever can be counted immediately.

The planning distinction dentsu wants to replace the old ones with is skippable versus non-skippable - not linear versus digital, big screen versus small screen, or platform A versus platform B.

The advertising Doom Loop: budgets shift to "performance" formats, advertising playback plateaus, teams optimise against faulty metrics, and the cycle repeats. Relentless optimisation for clicks and conversions diminishes effectiveness while eroding brand trust, and marketers default to short-term investment because it is what can be quantified immediately.

High demand power comes with nine times the volume share and four times the odds of growing

Kantar's Meaningful, Different, Salient framework, externally validated by the Marketing Accountability Standards Board, finds that brands with high demand power have 9x higher volume share and are 4x more likely to grow.

42% of finance leaders cannot measure the long-term impact of brand

Brand building is marketers' number one priority for 2025, yet it does not appear in finance leaders' top five, and 42% of finance leaders say they struggle to measure the long-term impact of brand building.

Loop Earplugs doubled orders in a quarter by tripling creative testing volume

Loop Earplugs, working with agency TRGT, went from 5 to 20 A/B tests and from 45 to 120 creatives tested in three months, plus a conversion lift study. Revenue rose 119% and orders doubled quarter on quarter.

Storytelling is instrumental to performance, not the opposite of it. The idea that brand is storytelling and performance is not comes from the banner-ad and DCO era; short-form video performance advertising is now one of the best ways to tell a brand's story at scale to exactly the right people, and it is how younger audiences form affinity with brands.

Pilot, Eucalyptus's men's health brand, put up a provocative billboard that was pulled after roughly a day. The buy itself was cheap and modest, but the photo of it ran through performance and organic channels, the takedown became a story about men's health, and the attention outlived the placement many times over.

Eucalyptus stopped trusting in-platform conversion numbers and rallied its team around creative efficiency instead: scaling spend into ads with strong hook rates, hold rates, click-through rates and top-of-funnel conversion, then triangulating with backend results to confirm that more efficient creative really produces more conversions.

Eucalyptus sets a standard media mix each month but treats it as a starting point: everyone who touches a media dollar owns the whole budget pool, a weekly all-brands review asks where money went, whether it was incremental, and where to pull from, and every month a no-strings pool is reserved for a new channel or a creative or brand bet that is allowed to flop.

For brand building without a Super Bowl budget, Eucalyptus finds broadcaster video on demand (BVOD) and YouTube the most reliable placements: both grow awareness and the customers they bring in convert at a rate that works for the business.

ElevenLabs ran dubbed video ads with the on-screen text still in English as a deliberate cheap test - it worked, the lift study confirmed incrementality, and only then did they request budget for full localization (lip sync, rewritten text).

Set per-language guardrails before AI localization: lock product and brand names, define tone per market. ElevenLabs' own engine translated their brand 'ElevenAgents' into Korean before guardrails caught it.

Video ad signal arrives slowly and disguised: clicking a YouTube ad pauses the video, so conversions surface later as brand search and direct traffic. Without a lift study you'll conclude the ad failed while it quietly fills other channels.

ElevenLabs' Ads Engine started as an internal tool (client of one) solving their own localization pain, then became a product - dogfooding as product strategy.

The practical case for brand awareness as a metric: it is not what people saw or what they clicked, it is what they actively went looking for. As the report puts it, when people know and trust a brand they do not search the category, they search for the brand - and that is demand rather than a vanity number.

For brands with established fame the job changes rather than ends: from getting known to staying front of mind, defending salience and keeping memory structures fresh at buying moments. The operational trigger is specific - when Share of Search stops rising in line with awareness, shift investment down the funnel toward consideration and preference.

Google's planning benchmarks are 40% weekly on-target reach and 2.7x weekly frequency

Brands with higher awareness were more likely to be following established media planning principles, which in turn correlated with greater Share of Search. Google's two published numbers for these Golden Rules: a minimum 40% weekly on-target reach for peak campaign effectiveness, and a median optimal weekly frequency of 2.7x on YouTube.

Google puts its own search volume at over five trillion queries a year and says AI is growing it

Google states it processes over 5 trillion searches annually, that commercial queries are rising year over year even at that scale, and that AI features like AI Overviews and AI Mode are growing the overall search pie rather than shrinking it. This is Google's own claim inside a Google co-authored report.

Vaseline's creator-led hacks campaign lifted ad recall 15 points without a traditional brand campaign

Vaseline built a brand campaign entirely out of creator-made 'hacks' - the spice hack, the runner hack, the hair dye hack - run as partnership ads rather than as a traditional brand campaign. It lifted ad recall by 15.0 points and consideration by 4.2 points, and took a Cannes Lions Titanium plus two Grand Prix.

Optimising ads to in-store as well as online sales multiplies incremental purchases

Ads optimised against both online and in-store sales drove 2.8x more incremental purchases than the same ads optimised for web sales only.

When an algorithm mediates discovery, the machine is the first buyer to convince. A product whose data, imagery and attributes are illegible to a recommendation model never reaches the person who would have bought it - if the machine cannot see it, the human cannot buy it.

Business messaging is opened three times more often than email

Messages sent through WhatsApp are opened at an average rate of 79%, against 26% for email.

Where select short-form video formats win is efficiency, not effect size: they achieve a lower cost per 1% of long-term sales lift than Linear TV. The size of the lift and the price of the lift are two different arguments, and short form only wins the second one.

Half of marketing leaders cannot tell whether CTV works as well as Linear TV

49% of global marketing leaders say they find it challenging to assess whether new video formats like Connected TV can deliver as effectively as Linear TV.

The reason voluntary attention pays is agency: a viewer who keeps watching a skippable ad is choosing to give it attention. dentsu argues this voluntary attention lives in the same space as earned attention, and that this is where brand equity actually gets built.

Buying past thirty seconds buys length the attention curve no longer pays for

If a consumer watches about half an ad's duration most of the time, even on non-skippable formats, then the value of ads longer than 30 seconds should be questioned - you are buying length that the attention curve has already stopped rewarding.

HexClad noticed in its Meta account that ads featuring Gordon Ramsay were its top performers, turned that insight into a formal partnership (Ramsay later invested in the company), scaled it up to a 2025 Super Bowl ad, and then cut the Super Bowl spot back into a range of Facebook ads so the celebrity credibility could be targeted at cooking fans and Ramsay fans.

Creative graduation: use performance channels, where data is cheap and immediate, to test as many angles, creators and messages as possible; graduate the winners to a higher-production YouTube ad, another channel with reliable metrics; once it works there, commit the budget to TV and extend the same message into PR, the website and organic. Every step adds confidence, and the result is one unified message across channels.

A pre-purchase 'how did you hear about us?' survey lets Eucalyptus pair each referral channel (TV, radio, out-of-home) with its spend, conversion and LTV, turning brand channels into something that can be compared on efficiency. It is imperfect self-reported data, so it is triangulated with brand tracking, ad-account creative data and GA first-party data.

Eucalyptus replaced its siloed brand and performance teams with a full-stack team per brand (growth or creative lead, growth marketer, creatives, producer) that owns its budget and targets across everything from TV to Meta ads. Hire storytellers in every role, not just copywriters, and cover three ownerships: an allocator of capital, an allocator of ideas and an allocator of time, because a TVC can consume the whole team for weeks and kill performance creative velocity.

Localized non-English ad campaigns convert 17.6% better than English ones, worth $3.8M in incremental value at a 7.16 ROAS.

ElevenLabs' non-English ad campaigns convert 17.6% better than English ones, generating $3.8M in incremental conversion value at a 7.16 ROAS - from the same winning ads localized into 50+ markets, not from new creative.

The proof-point ladder for winning budget: make a strategic bet where the data says customers are, ship the cheapest testable version, run a lift study to prove incrementality, then trade the case study for resources. Repeat with bigger asks.

Run ad translations through multiple LLMs cross-checking each other rather than a single model - slower on purpose, consistency over speed.

Before localizing ads into a language, check operational rigor: if you run Hindi ads without Hindi support, your support tickets arrive in Hindi. Demand you can't serve isn't growth.

Localize where the journey is repetitive, searchable and measurable first; use human review exports (native speakers verifying translated campaigns) as the escape hatch before spend goes live.

The study's chart is built to be misread. Brand Awareness Increase sits on the vertical axis and Share of Search Gain on the horizontal, which inverts the usual reading: the steep line belonging to high-awareness brands looks like the winner and is the opposite. A shallow line means each point of awareness bought more search demand.

Share of Search moves 6 to 12 months before market share does

Share of Search works as a leading indicator rather than a report card. The study cites Les Binet's finding that Share of Search correlates strongly with market share, and that changes in Share of Search often precede shifts in market share by 6 to 12 months.

For brands at the lower to mid end of the awareness scale, every point of awareness growth translates into measurable behavioural change in search and ultimately in market share. Growth-stage and disruptor brands have the most to gain, which makes brand investment at that stage one of the most commercially efficient moves available.

The argument for why AI search raises the stakes rather than lowering them: AI search surfaces one or two answers rather than a page of options. As the report puts it, there is no second page and no consolation prize for being the second-most-familiar brand in the category.

Attention accumulated across repeated short exposures converts better than the same seconds in one impression

Attention accumulated across repeated short exposures outperforms the same attention delivered in one sitting. Viewers reaching five cumulative attentive seconds in a week converted 21% better than viewers who got five attentive seconds in a single impression; at ten seconds the advantage was 13%.

Most retail sales still happen in store and stay invisible to digital optimisation

73% of retail sales still happen in store, and the forecast for 2029 is 70%. The larger half of retail remains largely invisible to digital performance optimisation.

Half of all media time is now spent with user-generated content

50% of media time is now spent with user-generated content. A brand with no face in creator content is absent from half the market's media consumption.

dentsu's own caveat on its CTV finding, printed beside it: "While CTV is delivering more than it used to, Linear TV is still guaranteeing high brand effects." The shift is in what CTV can now be trusted to do, not in Linear TV becoming replaceable.

Second for second, voluntary attention works harder than forced attention. Non-skippable ads deliver impact even at low attention levels - sound on, watched longer, and more expensive per impression - but skippable formats, once viewed for more than one to two seconds, surpass them.

Attention stops paying its way after about twenty seconds

More attention does produce greater brand-building effects, but the return flattens: after roughly 20 attentive seconds, additional attention delivers little extra impact. It contradicts the assumption that more attention is always better.

Brands do not plan against a composite score, so The Brand Reset breaks its findings out for Meaningful, Different and Salient separately, each with its own attention curve. That lets a planner match media and creative to the actual objective - awareness, consideration or differentiation - rather than to an average of all three.

The Brand Reset is agency research whose platform partners are the platforms being tested - Snap, Pinterest, Teads, Spotify and TikTok among the ten. The methodology is serious, the consultants are independent names, and the commercial interest is real. Cite the findings with the partnership named.

Brand adds commercial value in three ways a CFO already tracks: customer acquisition gets easier (trusted brands convert faster and cheaper, less discounting, better CAC:LTV), more value is realised per customer over time (pricing power, repeat purchase, margin), and brand plus community becomes a moat that is hard to replicate and insulates against downturns.

Digital performance channels take 73% of global ad spend

Digital performance channels account for 73% of global ad spend, and it is performance creative that built and scaled today's most influential challenger brands such as Gymshark, Hismile, Dollar Shave Club and Wild.

A $2,000 billboard bought for the photo, not the footfall

MagicBrief bought a Sydney billboard for roughly $2,000 not for the passers-by in Surry Hills but so it could run a photo of the billboard as a Facebook ad, carrying the credibility of out-of-home with the reach and targeting of performance media.

George Howes' 15 elements of a high-performance creative team come down to a tight feedback loop between media buyers and creatives, a culture that tests instead of debating, metrics owned by the whole team, a shared and regularly updated messaging matrix, a shared understanding of what drove past wins, the ability to iterate winners by going deeper rather than sideways into replicas, following the eye of the cultural storm, clean repeatable processes, acting on competitor formats, and a culture that keeps taking bets with outsized returns.

Juniper, Eucalyptus's medical weight loss brand, started from the insight that people saw medical weight loss as cheating. It tested hundreds of Meta ads arguing 'medical weight loss is not cheating', graduated the winning language into a YouTube explainer, then a TV campaign, and finally a national white paper and PR campaign on sentiment around medical weight loss, all carrying the same message.

Self-reported attribution under-reports brand channels. During a TV campaign, Eucalyptus sees a sudden spike in customers who say they heard about the brand on Meta or Instagram even though nothing changed on those platforms: the TV ad pushed them over the edge, but they remember the channel where they had already seen the brand. Read survey data as mental availability, not as a direct cause.

Brand creative should not be judged by immediate sales alone, but it can be measured in the short term: look for short-term lift in mental availability (brand recall, consideration, salience), brand searches, direct traffic and top-of-funnel growth by referral channel, and compare awareness and consideration movement against competitors to see whether you are outpacing the market.

Translation is not localization: what works in the US won't work in Japan or India. Beyond words, brands must adapt tone, creative, colors and site experience per market - ElevenLabs saw Portuguese and Spanish succeed while the same playbook failed in Japanese.

Naive AI translation in spreadsheets fails predictably: it exceeds character limits it was explicitly given, translates the same headline three different ways, and the only quality check is back-translation. At hundreds of thousands of ads this means days of manual verification per language.

Voice consistency is a brand asset in dubbing: iconic voices (Hasbro characters, celebrity talent) must keep their tone, pacing and character across languages - a key requirement for AI dubbing at brand scale.

Ad platforms are getting hungrier for creative volume - Meta's Andromeda era demands more variants than in-house teams can produce, which is the core argument for AI-assisted creative localization at scale.

Brand marketers have a structural disadvantage that performance marketers do not: their most important work is the hardest to prove. So the metrics that win budget meetings end up being the ones easiest to measure - impressions, clicks, reach - rather than the ones that indicate whether anything is being built.

For mature high-awareness brands the growth is at the fringes rather than in the middle: new category entrants, lapsed users and low-frequency buyers. The recommended combination is distinctive assets and category cues paired with expansive targeting and scaled reach.

The study deliberately uses traditional search data rather than AI-powered discovery, because traditional search provides a long, consistent, measurable record of how consumer intent behaves over time. The authors state plainly that the data to quantify the AI shift at scale is still maturing.

Google co-authored a study concluding that Google Search volume is a good proxy for brand health and that Google's own YouTube planning rules correlate with success. The base is stated and the method is transparent, and the commercial interest is real. Cite the finding with the authorship attached.