Current answers to the marketing questions consumer businesses ask. Sign in free to read them all.1,898 sources, at least 3 behind every answer.

Season or you, part two

Black Friday is 8 days out and ad prices are climbing right on schedule, so this issue opens with how much of your CPM rise is the market and how much is you. After that: AppLovin's comeback tour, Meta pressing pause on its new terms, a Google sale setting with a new answer behind it and two new answers in the book.

By Scott Stroud, Editor

Cartoon: a marketer finds her receipt far longer than the market's, then pastes a fresh poster over a faded billboard

Tuesday's issue gave you a test for whether rising costs are the season or your account. Now there's a market line to hold it against. The incrementality testing seller Haus says Meta CPMs for its brands have run about 11% higher since October than the same brands paid over the summer, and AppLovin's about a quarter higher (Olivia Kory). Our answer on lowering high Meta CPMs starts exactly there. Compare your rise with the market first, using Meta's own price per ad, which was 23% higher than a year earlier in the US and Canada in Q3 2026. If yours rose more, fix it with new and distinct creative, broader targeting and placements, fewer and better-funded ad sets and clean tracking, and judge every change on cost per purchase rather than CPM.

Andrew Foxwell offers one reason a CPM climbs alone. He says Meta's Andromeda 2.0 groups ads by concept, reading setting, format and audio, so look-alike ads cluster together, and an ad that never got spend has a similarity problem rather than fatigue (Foxwell Digital). His 20-minute audit tags every active ad by format, tone and angle, and flags any format at 50% or more of active ads, any tone at 40% or more and any angle at 30% or more. He blames that kind of sameness for higher CPMs more often than competition. The thresholds are his rule of thumb, with no data behind them, but our answer on what counts as a new creative draws the same line: new copy, a new crop or a new opening line over the same footage is a variant, and Meta likely treats it as the same ad. A voiceover swap the week before Black Friday is a costume change, not a new cast member.

On timing, Cody Plofker says push spend harder in the run-up, because pre-sale spend keeps paying off once the sale opens and the last day has little left to give (Cody Plofker). Our answer on budgets for seasonal peaks holds or raises spend ahead of the sale too. Just expect those run-up impressions to cost more than they did in July.

AppLovin's comeback tour

The same Haus post refreshed its AppLovin numbers. Since July, 79% of its AppLovin tests beat the same brand's other channels, against under half in the spring, and brands with a clearly positive test raised AppLovin from about 6% to about 9% of paid budget while everyone else cut it from about 4% to about 2% (Olivia Kory). Haus gave no test or brand count, and it's a testing company describing its own tests. Our answer on how stable AppLovin is for ecommerce told you to expect swings from quarter to quarter, and under half to 79% is a swing. The advice holds: run it beside Meta and Google, never in place of either, and rerun a holdout as you scale.

TV corner

  • Kory also says to compare connected TV with the marginal dollar in Meta or Google, not their average, and on that basis Haus tests put connected TV about even, at 0.98 times (Olivia Kory). That's the logic in our answer on when connected TV makes sense, which adds TV when Meta's next dollar starts earning less.
  • Mike Shields of Next in Media, citing people familiar with the talks, reports Meta is discussing an "extend to CTV" option in its ad-buying tools with TV makers and free streaming platforms (AdExchanger). Talks only, no product.
  • Our answer on connected TV or YouTube starts with YouTube for now.
  • Amazon and Philips say Philips Sonicare's streaming buys through Amazon Ads Agent brought 24% incremental reach and a 75% lower customer acquisition cost against shoppers who saw only Sponsored Ads (AdExchanger). That's Amazon grading its own homework with an exposed-against-unexposed comparison.
  • Our answer on Prime Video and Amazon streaming ads wants a holdout before those numbers count.

Meta notes

  • Callback to OctoberMeta's revised Commercial Terms, the ones that let its crawlers read your site, stay on hold. Meta says it will consult publishers over the next few months, take revised terms to stakeholders and regulators in early 2027, notify account holders before anything applies and that its crawlers follow robots.txt (Meta).
  • Callback to OctoberOur answer on letting Meta's crawlers read your pages holds: let the ads crawler in and set the AI-training crawler by its own robots.txt rule.
  • AdExchanger reports Meta is running campaigns to get people to install its Muse agent as a Chrome extension, which AdExchanger says could help Muse get past the bot detection retailers such as Amazon and eBay use (AdExchanger). Meanwhile Loblaw became the first Canadian retailer connected to Muse, with checkout inside the agent from November 18 and no usage figures (Meta). An agent inside a shopper's own browser looks like a normal visitor, which is why our answer on AI agents in conversion data bids on backend-confirmed purchases rather than checkout starts.
  • The agency Common Thread Collective says that across hundreds of brands, Meta's 7-day click reporting understates its business impact by about 20%, rising to 1.4 times for brands that also sell on Amazon (Common Thread Collective).
  • Our answer on crediting Meta for Amazon and retail sales counts that spillover. The same post says each $1 of TikTok Shop revenue comes with about $0.53 more captured elsewhere, from $0.20 to $0.93, which fits our answer on how TikTok Shop affects site sales.
  • Lorenzo Pravata says Triple Whale's Moby, the analytics seller's AI tool, puts the median Meta cost per purchase for pet brands with order value under $100 at $36.67, with the top quarter under about $25 (Lorenzo Pravata).
  • Our answer on benchmarking your CPMs and costs treats any benchmark as a starting point against your own history.
  • Sarah Levinger lets a new ad run 7 to 10 days before reading it (Sarah Levinger), which matches our answer on judging a creative test. Do the math on a launch today and you're reading it on Black Friday, which is a fine day to learn and a bad day to decide.

Google

  • The Performance Max setup screen now shows Channel Prioritization and Promo Mode under Budget. Neither is new, they just moved into the flow (Search Engine Roundtable). Good timing for a new answer. Our answer on Google promotion mode for sales, new on Wednesday, says use it for a sale of your own running 3 to 14 days on a standard Search or Performance Max campaign with its own Target ROAS, when margin can absorb a lower ROAS for more sales. Agree the lowest ROAS you'll accept first, add budget so the campaign isn't limited by it and leave it alone until it reverts. Skip it for flash sales under a day and for portfolio strategies or shared budgets. It also says try it on a smaller sale before Black Friday, and with 8 days left, that ship has mostly sailed.
  • Google's November Demand Gen update makes seasonality adjustments available in Demand Gen with Target CPA or Target ROAS (Google). Our answer on seasonality adjustments for short promotions keeps them for short sales Google can't see coming, and Black Friday isn't one.
  • Callback to Wednesday's plumbers. Our answer on Local Services Ads cost and value changed its call on Wednesday: with manual lead disputes gone once your account moves into Google Ads, it now says turn off services you don't offer and review leads on the Leads page, in place of counting on credits.
  • The Google Ads agency Echelonn says Google's AI review summaries are now showing in Search ads for one of its clients, positive only, not editable and shown only when Google predicts they help (Echelonn). Our answer on Google's review summaries in Search ads already says the fix is more and better reviews, since there's nothing to edit.
  • Common Thread Collective says audit Search campaigns that moved to AI Max now: filter the search terms report for AI Max matches to catch competitor, brand and informational queries, and add URL exclusions for pages with no path to purchase (Common Thread Collective). Travel brands get their own version, as Google published seven help pages on AI Max for Search campaigns for travel, where Hotel campaigns are headed (Google Ads Help). Our answer on turning on AI Max for Search, Wednesday's lead, holds.
  • Video campaign groups in Google Ads can reportedly mix reservation buys, such as YouTube Select and the CPM Masthead, with auction campaigns in one group (PPC News Feed). Google's own note isn't in our sources yet. Our answer on YouTube ad costs and budget starts in auction, so this matters mostly to bigger budgets.
  • Menachem Ani runs Performance Max feed-only, with asset optimization off, no assets beyond the feed and no audience signals or search themes, to push it toward Shopping (Menachem Ani). That's our answer on feed-only Performance Max. In another account of his, a dedicated landing page converted at 3% against 0.25 to 0.5% for the homepage on the same traffic (Menachem Ani), one more vote for our answer on where to send cold traffic.

Search and AI

  • Google AI Overviews on desktop reportedly show all three citations at full strength, in place of one clear card and two faded ones, with no word on whether it's a full rollout (PPC Land).
  • Our answer on AI Overviews and clicks judges search on leads and sales, so watch whether the second and third spots start sending any.
  • Two Northwestern professors, writing in Adweek and citing the SEO software company BrightEdge, say YouTube gets nearly 30% of citations in Google's AI Overviews (PPC Land). BrightEdge's sample isn't given.
  • Our answer on video for SEO covers when video earns the work.
  • Google's Search Console API now returns the branded and non-branded split, multimodal search data and multi-country filters that were only in the interface (Search Engine Land).
  • Our answer on monthly SEO metrics reports non-brand clicks on their own line, and now your dashboard can too. Joy Hawkins of Sterling Sky also says she still sees fake clicks in Search Console for some sites, visible as very high click-through rates on keywords where the site doesn't rank near the top (Joy Hawkins).
  • Our answer on using Search Console data says check odd keywords at page level before you celebrate.
  • Callback to Wednesday's Google-Agent guide. Blend found bot protection blocked its AI shopping agent at 29 of 200 retailers (PPC Land).
  • Our answer on bot blocking and Google Shopping says check your CDN and firewall settings, since robots.txt doesn't decide it.
  • The CEO of Stacker, which distributes brand stories to publishers, says brands are hiring former journalists because AI search tends to cite pages with original data (Marketing Brew).
  • Our answer on paying for digital PR pays only when you have data or a story, so the journalist needs something to write about.

Elsewhere

  • The mix-model seller Recast says its new add-on estimates customer value by acquisition source, and on client data the spread ran from two partner sources at about 1.2 to 1.3 times organic customers' value to every paid channel within about 7% of organic (Recast).
  • Our answer on lifetime value by acquisition channel measures it on first-order cohorts by channel.
  • In DoubleVerify's survey of 2,020 marketing decision-makers in 25 countries, 55% measured campaigns with platform reporting, 48% with mix modeling and 35% with last click, and incrementality testing wasn't offered as an answer (PPC Land).
  • Our answer on choosing an attribution approach would have checked the box they left off.
  • A copywriter reports 1.8 to 2.7% placed-order rates on plain-text founder thank-you emails after purchase, against 0.3% on branded ones, in his own clients' accounts (Direct to Consumer). No sample size.
  • Our answer on crediting email and SMS fairly would want to know how many of those orders were coming anyway.

New in the book

Two new answers on Wednesday, one already in the Google section. The featured one: our answer on product photos and returns says read return reasons product by product, and where returns say an item wasn't as pictured for color, size or look, reshoot it true to color, add scale and close-up shots, match the description and compare that product's returns on orders placed after the change, allowing 4 to 8 weeks. Black Friday returns arrive in December, so this one is for your January self.

Thanksgiving is in 7 days. Black Friday is in 8 days.

Every fact links to its source. Opinions are the writer’s, built from the sources linked beside them.

Every issue