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.

The fine print, version 27

Meta shipped a new version of its ads API, and three of its changes switch on more automation for ads built through a tool instead of by hand in Ads Manager. Also: a new call on Meta's new-customer setting, Google's Promotion Mode opening up before Black Friday, and a new Google answer for anyone whose shoppers now check out on Google itself.

By Scott Stroud, Editor

A big power switch, flipped on in red, sitting inside a laboratory beaker

Meta's Marketing API v27.0 went live November 4. It's the plumbing that ad tools, bulk uploaders and agency dashboards use, so these changes can reach you without anyone on your team opening a developer page (Meta for Developers). Product showcase, which shows matching catalog products with your ad, may switch on for new API-built ads that carry no showcase setting, "only when it has high confidence that a specific product set in your catalog matches the ad". New Sales ad sets that target specific mobile devices or operating systems get "Reach more people across other mobile devices" turned on unless the tool sets the device field to 0. And lookalike country settings now get rejected, because Meta says they "have no effect" and lookalikes deliver wherever the ad set targets. Existing ad sets and audiences stay as they are, and older API versions follow in about 90 days. PPC Land points out that Meta gives no performance data for the new defaults (PPC Land).

My advice: ask whoever runs your ad tool which API version it's on and what it sends for these fields, because "we didn't touch anything" is exactly how a default sneaks in. Our answer on Meta's creative enhancements now notes Product showcase as one more feature that can arrive switched on, and its call stays put: decide feature by feature, leave format features on, and fence anything that rewrites your words or alters your product images until your own Creative report shows it beating your originals. Confidence is low, since the only performance figures come from Meta grading its own tools.

The device default sits fine with our answer on restricting placements, ages or genders, which says don't hard-restrict a group and bid it down with a value rule once a breakdown shows a real problem. The lookalike change is housekeeping that matches our answer on broad targeting: go broad, and treat lookalikes as suggestions. If you sell by the lead, v27 also adds phone code verification to lead forms, as required, optional (eligible accounts only) or off (Meta for Developers). Our answer on cutting junk leads says filter junk where it enters and judge every change on qualified leads, not raw lead count, which is how I'd judge this one too.

New customers, at a price

Our answer on Meta's "Acquire new customers" setting is new. Turn it on only in ad sets whose job is first-time buyers, test it on a duplicate ad set beside your current setup for at least two weeks with today's cost per new customer written down first, and expect a higher cost per result. Judge it on first-ever orders counted in your store or CRM, not on Meta's purchase count. The cost warning comes from Madgicx, which sells ad tools, reading Meta's interface, because we couldn't find a Meta help page for the setting (Madgicx). In one Pilothouse account on the setting, purchases cost $72.63 against $38.44 for engaged audiences (Pilothouse). The dissent runs both ways: CTC wants it in every prospecting ad set for Black Friday (CTC), and Ben Heath says use it only when spend is piling onto existing customers (Heath). Medium confidence on the test, low on whether it lowers what a new customer costs.

Shop talk

  • Jon Loomer says on his podcast, "Only consider a second ad set if you're generating at least 100 conversions per week" (Loomer).
  • Our answer on how many ad sets to run sets the bar at about 50 a week per ad set, so Loomer is the stricter coach. Phil Kiel, who sells Meta management, sees the cost in audits: setups so tangled "that even if they launched a new banger, it wouldn't float to the top" (Kiel).
  • Jordan West relays a 5-week Fenty Beauty geo-lift study run with WorkMagic, which sells measurement, that found "6.7X in incremental sales BEYOND purchases made directly on TikTok Shop", with Fenty's own site taking nearly twice the halo of all other channels combined (West). One brand, a seller's study.
  • Our answer on how TikTok Shop affects site sales already allows a site lift bigger than the Shop's own gain, and Fenty is a loud example of it.
  • Olivia Kory of Haus, which sells incrementality tests, says a study comparing ChatGPT ad returns with search ads ran on last click, so "the direction of the insight is interesting but the magnitude is overstated" (Kory).
  • Our answer on ads inside AI assistants took in a study of that kind, where a paid ChatGPT visit earned far less than a paid Google search visit, and kept its call: test ChatGPT only with money you can treat as research, judged with a regional holdout on your own sales.
  • Barry Hott wants you off hook and hold rates"Rates don't account for audience relevance, quality, or costs", so he reads cost per video play and cost per ThruPlay (Hott). Opinion, no data, and our answer on judging a creative test agrees that hook rate is no verdict; it confirms on cost per purchase or ROAS over at least 7 days.

Google

Promotion Mode now has its own help page as an open beta, so you no longer go through your account team. On Search and Performance Max campaigns with campaign-level Target ROAS, it lowers the target by a tolerance of 5% to 90% and can add daily budget for a 3- to 14-day promotion, then reverts on its own. It doesn't work with portfolio bid strategies or shared budgets, and it stacks on any Smart Bidding Exploration tolerance, so 20% plus 20% makes 40% (Google Ads Help). Our answer on Google budgets around sales peaks wants Black Friday budgets at about 3 times expected spend and ROAS targets lowered only in gradual steps a few days ahead, so if you try it, I'd keep the tolerance small.

If AI Max matches your searches, check your reports. Google added a "Search terms match type" field that includes AI Max and renamed the old one "Search terms match type (Keywords)", which Mike Ryan says covers only standard match types in Search campaigns (Search Engine Roundtable). For now the new field shows AI Max only when the Search term column is in the report, and Google says a fix is coming, with no date (Google's Ads Liaison). Our answer on AI Max keeps it on a test judged against total sales, and a report that drops AI Max matches makes that test harder to read.

For the phone-call crowd: Google's Data Manager API now takes call conversions and flags which records in an upload failed, since "a successful ingestion request doesn't guarantee that every record was processed" (Google Ads Developer Blog). Menachem Ani, who runs an agency, adds a cautionary tale: when a Google Business Profile phone number changed on a lead-gen account, call conversions quietly stopped, Smart Bidding sagged, and his team bid manually on clicks for 7 to 10 days (Ani). Our answer on tracking phone sales covers the setup, and Ani's fix costs nothing: test-call your own ads.

Two new Google answers. Sales that check out on Google: count Buy now and Universal Cart orders from your own order system, since your site tags never see them, and judge Shopping and Performance Max on total store sales, because Google hasn't documented how a Buy now purchase after an ad click is credited in Google Ads. In the feed, mark only products Google's checkout supports and keep return and support details complete (Google). Low confidence on the credit, medium on the rest.

And Google ad schedules: keep the default all-day schedule unless nobody can serve customers at some hours, and skip tuning schedule bid adjustments, which Target ROAS and Target CPA ignore. Since June 2026, a schedule that switches off whole days paces toward 30.4 times the average daily budget for the month, up to 2 times on any active day; Google's own example has a campaign scheduled for 20 days at $100 a day pacing toward $3,040 instead of $2,000 (Google Ads Help). Weekday-only advertisers, check your monthly bill.

New in the book

One more new answer, for smaller budgets. Our answer on marketing mix models now has a version for brands under about $200K a month: skip a paid mix model, judge Meta on its own reporting beside your store sales and GA4, and run periodic lift tests, starting with Meta's free Conversion Lift once you qualify. Flighted, an agency, dissents with a mix model from $50K a month across 3 or more channels (Flighted).

Meta's on-device counting starts November 12, in 7 days. Black Friday is 22 days out.

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

Every issue