Let's Talk
Let's Talk
JOURNAL
Industry dispatches, playbooks, and account teardowns from the team running paid-media at scale — what shipped, what broke, what we're testing next.
Ask what a good Facebook ad frequency is and you'll get a number: 2.5, 3, maybe 3 to 4. We traced each one back to its source. None led to published evidence for sales campaigns on Meta, and the famous 3 started life as a minimum, not a ceiling. What Meta says is far vaguer than the rules of thumb, and the controlled research points somewhere else: to the trend, not the number.
OpenOn October 6, 2026, Meta made Customer Lifecycle Strategy available to all advertisers. Mark a Sales ad set as acquisition and it stops spending on the customers you exclude. It can only exclude people it can match, though, and it will probably make your ROAS look worse. Judge it on new-customer CAC instead, counted in your store.
OpenMeta has told advertisers that starting October 28, 2026 it will add product browsing to their existing Facebook and Instagram ads unless they opt out. A tap on your ad may then open a product screen inside Meta's apps before anyone reaches your site. The pitch is 3% more conversions, a number that leaves out most of what you'd need to judge it. Here's how to turn it off ad by ad, and how to test it properly once the holidays are over.
OpenMeta's average price per ad rose 12% year over year in both the first and second quarters of 2026, and brands on Triple Whale paid $22.26 per thousand impressions over last year's BFCM. Black Friday lands on November 27. Before budgets go up, work out the highest CPM your margin can carry, because the weekend is a race between what an impression costs and how often it turns into an order.
OpenChatGPT ads went self-serve on May 5 with no minimum spend, and OpenAI suggests starting bids of $3 to $5 a click. For a store, that price is most of the story. At a 40% contribution margin and a 4% conversion rate, a $4 click pays back on the first order only if the basket is around $250. The first named tests paid between $4.41 and $9.89 a click.
OpenIn late August a new column appeared in Ads Manager: Creative diversity, rated Low, Medium or High for each ad set and marked as estimated and in development. It's Meta telling you how many distinct ads it thinks you run, and the answer is often lower than your ad count. A Low rating is fixable without pausing the ads that pay.
OpenThe most common discount on Shopify last Black Friday was 20% off. On a 35% gross margin, that code means you need 133% more orders just to make the same profit. Most brands get 30% more and call the sale a win. Here is the math the promo calendar never shows you, and what to run instead.
OpenWe run paid media for a living, so this is an odd article for us to publish. Clients fire agencies over delivery and value; agencies think they get fired over budget cuts. That gap is where bad relationships hide for years. Here are the twelve questions that close it, including the ones you should ask us.
OpenTake conversion rate from 2% to 3% and your cost per customer falls by a third, with no change to bids, budgets, or creative. Almost nobody spends a week on it. Here are the 2026 benchmarks by industry, device, and channel, the levers with real evidence behind them, and the math that makes CRO the cheapest ROAS you will ever buy.
OpenGoogle clicks cost about four times more than Meta clicks and convert about four times better. That is not a coincidence; it is the same customer at two different moments. The real question is not which platform wins. It is which job each one does in your funnel, and in what order to buy them.
OpenThe standard answer is a percentage of revenue: 10% to 20% for ecommerce, more if you are pushing growth. It is a decent sanity check and a terrible way to set a budget, because the right spend comes from your margins and payback, not from what other companies do. Here is the backwards method.
OpenAverage CPC is up to $1.72, CPMs jumped 20% to about $14, and the median ecommerce ROAS sits near 1.86. Useful numbers. Also the fastest way to make a bad decision, because a benchmark is not a target. Here are the 2026 figures and how to read them against your own math.
OpenGoogle's AI Max and broad match infer intent, which means they quietly wander into your brand-name auctions and bid against your own brand campaign. You pay more for clicks you would have won cheaply, and your reported ROAS inflates. Here is how to catch it and claw the margin back.
OpenMore buyers now ask ChatGPT or Perplexity for the best product before they ever open Google. If the model does not name you in the answer, you lost the sale before the click. Here is how a DTC brand earns those recommendations.
OpenPlatform ROAS answers a question you did not ask. It tells you how many sales the platform took credit for. The question that decides whether to keep spending is different: how many of those sales would have happened anyway, with no ad at all?
OpenEvery ad platform claims the same sale. Meta counts it, Google counts it, TikTok counts it, and your three dashboards add up to more revenue than your bank received. There is one number nothing can double-count, because it starts from the money that landed.
OpenMarketing mix modeling used to be a thing only big brands could afford. Two things changed in 2026: privacy broke last-click for everyone, and the platforms open-sourced their own modelling tools.
OpenMeta switched on Threads ads worldwide in January 2026, and right now the placement is cheap because few advertisers have shown up. Early CPMs run well below Instagram. This is the short first-mover window, and here is how to use it without wasting budget.
OpenYou used to win by finding the right audience. Lookalikes, interest stacks, detailed targeting, that was the job. Then Advantage+ and broad match swallowed most of it. The machine picks the audience now, and it is better at it than you are.
OpenHere is a number that will mislead you faster than almost any other in a service business: cost per lead. A clinic runs a campaign, the form fills come in at $6 each, and everyone celebrates. Three weeks later the calendar is half empty and nobody can explain why a flood of cheap leads turned into a trickle of booked patients.
OpenMost brands now know their reported ROAS is inflated. We wrote the diagnosis in why your ROAS is lying to you, and you have probably felt it: the dashboard says 4x, the bank account disagrees.
OpenGoogle says AI Max delivers 14% more conversions at similar cost. Independent analysis of 250+ real campaigns shows ROAS outcomes ranging from +42% to -35% — same feature, wildly different results. The honest version of what AI Max does, where it works, where it fails, and how to test it without burning budget.
OpenMost businesses don't have an ads problem. They have a system problem. Somewhere between the click and the customer, 30-50% of every ad dollar disappears into operational gaps nobody flagged because nobody was looking. The seven most common leaks we find in real ad accounts — and exactly how to fix each one.
OpenYour dashboard says 4x ROAS. Your finance team says revenue is flat. Both numbers are technically correct — and that gap is where most growth strategies quietly fail. The full breakdown of why platform-reported ROAS inflates by 30-50%, and the math that actually tells you whether your ads make money.
Open