CAC is up for almost everyone, and most of it isn't your fault
If your cost to land a new customer has been creeping up over the last year and a half, welcome to the club. Pretty much everyone's in it. The squeeze is coming from four directions at once, and only one of them has anything to do with how well you're running things.
The auctions are packed. Every brand that survived the post-2021 shakeout is now bidding on the same Meta, Google and TikTok inventory. And the platforms' own AI buying (Advantage+, Performance Max, Smart+) has levelled out the advantage a sharp media buyer used to give you. When everyone's running the same optimiser, the price of a click just keeps climbing until it hits the brand with the fattest margin, not the smartest team.
The platforms see less. iOS privacy changes, cookie restrictions and consent pop-ups mean Meta and Google know far less about what happens after the click. Less signal means broader targeting, more wasted impressions, and a CAC that drifts up even if your offer hasn't changed a bit.
Discovery has moved house. A growing chunk of product research now happens inside ChatGPT, Gemini, Perplexity and the AI answers sitting at the top of Google. That traffic is hard to track, hard to buy and tends to show up further down the funnel. So your paid-social numbers can look rough while overall demand is actually fine.
Shoppers are twitchier about price. Higher living costs and a decade of training people to wait for the discount code have dragged conversion rates down and discount depth up. Your ad spend buys the same visitor it always did; that visitor just needs more convincing.
Here's where most brands go wrong: they treat CAC as one number and react to it as one number. Cut spend, swap agency, try a new channel. But CAC isn't one lever. It's what falls out of five or six separate stages, and the fix depends entirely on which stage has actually broken.
'Blended CAC is up' tells you nothing. Break it apart.
Blended CAC is just total acquisition spend divided by new customers. Great as a scoreboard, useless as a diagnosis, because a 20% jump can come from six different places that each need a different fix.
So write out the chain that turns a pound of spend into a new customer:
CAC = (Spend / Impressions) × (Impressions / Clicks) × (Clicks / Orders) × (Orders / New customers)In plain English: what you pay per impression (CPM), how often an impression turns into a visit (CTR), how often a visit turns into an order (CVR), and how many of those orders are actually new people rather than existing customers you've just paid to reach again.
Then look at the two numbers sitting either side of CAC and ask whether it's even a problem: average order value and 90-day repeat rate. If CAC went up 20% but first-order AOV went up 30%, you've got more efficient, not less.
Pull each of those ratios for the last six months, by channel, and look for the one that moved. You'll usually see one of three patterns:
- CPM up, everything else flat. The auction got pricier. That's an inventory-cost problem, so the fix is audience, channel mix or creative that earns a cheaper placement. Not landing pages.
- CTR down, CPM flat. Your ads stopped earning attention. Creative's gone stale, or the message doesn't match what people are looking for anymore.
- CVR down, traffic cost flat. The site, the offer or the traffic quality changed. This one often shows up after a platform switches on broad targeting by default and starts sending you colder visitors.
There's a sneaky fourth pattern hiding in that last ratio: more and more of your paid orders are coming from people who were going to buy anyway. If retargeting and branded search have grown faster than prospecting, your real new-customer CAC could be miles above the blended number.
You don't need a data warehouse for any of this. A spreadsheet with six months of platform spend, impressions, clicks, orders and a new-versus-returning split from your store will show you where the break is in an afternoon.
The five places the bottleneck usually hides
1. Audience: you're paying to reach the wrong people, or the same people twice. Platform automation now defaults to broad targeting, which is brilliant for some categories and quietly floods others with cold traffic. Check frequency on your prospecting campaigns (anything over 3 or 4 a week is fatigue territory), check what share of your 'new customer' conversions are actually repeat buyers, and check your exclusion lists are still syncing. The biggest audience win usually isn't finding new people. It's not paying for people you already own.
2. Creative: the ad has stopped earning its spot. On Meta and TikTok, creative is targeting. The algorithm finds the audience that reacts to the ad, so a weak ad doesn't just convert worse, it gets shown to worse people at a higher price. Signs you're here: CTR sliding for weeks, hook rate (3-second views over impressions) under 25% on video, and a creative library where the same three winners have been carrying the spend since spring. Brands that get out of this hole are shipping 20 to 50 new concepts a month, not three.
3. Landing and conversion: the site is leaking what the ads paid for. The ad promised one thing and the page delivered another. The product page takes four seconds to load on a mid-range Android. Checkout wants an account. Shipping cost pops up at the very last step. If mobile conversion is under 1.5% for a mid-price consumer product, that's nearly always a site problem, not a traffic problem. It's the cheapest bottleneck to fix and the one most teams skip, because it's less fun than media.
4. Economics: CAC is fine, the order's just too small. If first-order AOV and repeat rate have dropped, a steady CAC still gives you a worse business. Discount-led acquisition is usually the culprit: a 25% welcome offer trains a customer who only ever buys at 25% off. Look at CAC payback period rather than CAC on its own, and at how many new customers place a second order inside 90 days. Bundles, free-shipping thresholds and subscription prompts shift this faster than anything you can do in Ads Manager.
5. Measurement: the number itself is wrong. Platform-reported CAC and your real CAC parted ways years ago. Meta will claim conversions Google also claims; both will claim conversions that were going to happen regardless; neither will see the customer who read a ChatGPT recommendation and typed your URL. If you haven't checked your attribution against a holdout or geo test in the last year, you might be optimising toward fiction. Rule this one out first, because fixing the others on bad data just shuffles budget toward whichever platform lies most convincingly.
The diagnostic, stage by stage
One question and one metric per stage. Work down the table in order, stop at the first row that fails, and fix that before you touch anything below it.
| Stage | Question to ask | Metric that answers it | Warning sign |
|---|---|---|---|
| Measurement | Does platform-reported CAC match what the bank account says? | Platform CAC vs blended CAC (total spend / net-new customers) | Gap wider than 30% |
| Measurement | Would these sales have happened without the ads? | Incremental lift from a geo holdout or conversion-lift test | Lift under 20% on retargeting or branded search |
| Audience | Are we paying to reach people we already own? | Share of paid 'new' orders that are returning customers | Above 15% |
| Audience | Are we showing the same people the same ad too often? | Weekly frequency on prospecting | Above 3 to 4 |
| Creative | Is the ad still earning attention? | CTR trend by ad, 3-second hook rate on video | CTR down 20%+ in 8 weeks; hook rate under 25% |
| Creative | Are we refreshing fast enough? | Net-new concepts launched per month | Fewer than 10 |
| Landing | Does the page keep what the ad bought? | Landing page bounce rate, add-to-cart rate | Bounce above 60%; ATC under 8% |
| Landing | Is the site fast enough on the phones people actually use? | Largest Contentful Paint on mobile | Over 2.5 seconds |
| Conversion | Where in checkout do people leave? | Checkout completion rate, step-level drop-off | Completion under 50% |
| Economics | Is the first order big enough to justify the cost? | First-order AOV, CAC payback in days | Payback over 90 days without a subscription model |
| Economics | Do they come back? | 90-day repeat rate for new customers | Under 20% for a consumable; under 10% for durable |
These thresholds are rough and they shift with category, price point and margin. A £300 furniture brand and a £15 skincare brand will land on opposite sides of half of them. Use them to spot the row that looks most out of whack for your business, not as pass/fail marks.
Tools worth exploring in 2026/27, by bottleneck
Buy for the bottleneck you actually found, not for whatever category is loudest on LinkedIn this month. The free and native options come first in each group because they'll answer most of your questions if you're under a few million in revenue. The paid platforms start earning their keep once a channel is big enough that a 10% measurement error costs you more than the subscription does. Prices in these categories change monthly, so treat anything below as a starting point and check the vendor's page before you commit.
| Bottleneck | Tool | What it does for CAC | Fits |
|---|---|---|---|
| Measurement | Meta Conversion Lift, Google conversion lift | Free in-platform holdout tests that tell you how much of a channel's claimed revenue is actually incremental | Anyone spending real money on one platform; run this before you buy anything else |
| Measurement | Google Meridian | Open-source marketing mix model with geo-test calibration; free, but you'll need an analyst | Brands with a data person and a couple of years of spend history |
| Measurement | Triple Whale | Shopify-native attribution with a unified MTA, MMM and incrementality add-on, plus AI-referral tracking | Shopify brands from roughly $1M who want one dashboard |
| Measurement | Northbeam, Rockerbox | Machine-learning attribution checked against incrementality tests | Growth-stage DTC brands running several channels |
| Measurement | Haus, Measured, Recast, Sellforte | Dedicated geo-lift experimentation and MMM platforms | Mid-market and enterprise brands where measurement is worth its own budget line |
| Creative | Meta Ads Manager generative tools, Advantage+ creative | Free variations, background expansion, text-in-image and translation inside the platform; expanded again at Cannes 2026 | Every Meta advertiser; use it for volume, not for your hero concepts |
| Creative | Segwise | Tags creative elements, flags fatigue and generates new variants from the patterns already winning in your account | Brands with enough spend history for the patterns to mean something |
| Creative | Melius | AI-assisted image, video and audio creation, with concept development and campaign variants in one creative workspace | Teams that need to turn briefs into more creative concepts and assets to test |
| Creative and paid acquisition | Lapis | AI agents that create, launch and optimise paid campaigns, connecting ad variations with targeting, landing pages and performance feedback | Growth teams whose bottleneck spans creative testing and campaign execution |
| Creative | Atria | Searches the Meta ad library and turns competitor angles into briefs | Small teams who need faster briefing more than more rendering |
| Landing and conversion | Microsoft Clarity, Shopify's built-in analytics | Free heatmaps, session recordings and checkout funnel reports | Everyone; find the leak before you pay to test fixes |
| Landing and conversion | Shoplift | Shopify-native theme and template A/B tests with no flicker, from about $74 to $99 a month | Brands testing page structure and layout |
| Landing and conversion | Intelligems, ABConvert | Price, shipping-threshold, discount and offer tests measured in profit per visitor, not just conversion rate | Brands whose question is 'what should this cost' as much as 'what should this look like' |
| Landing and conversion | Visually.io, Convert, VWO | Broader CRO suites with personalisation, server-side tests and analytics | Teams running an always-on testing programme, not one-off fixes |
| Economics and retention | Klaviyo, your subscription app, Intelligems threshold tests | Welcome flows that don't lead with a discount, free-shipping thresholds set just above current AOV, subscribe-and-save prompts at first purchase | Brands whose CAC is fine but whose first order and repeat rate aren't |
| AI search and new channels | GA4 AI Assistant channel, Search Console AI reports, Bing Webmaster AI Performance | Free first-party reports showing AI-referred sessions and revenue, AI Overview impressions and Copilot citations | Every store; switch these on before spending a cent on tracking |
| AI search and new channels | Shopify Agentic Storefronts | Syndicates eligible catalogues to ChatGPT, Copilot, Gemini and Google AI Mode for free (live since March 2026) | Any Shopify merchant; turn it on and clean up your product data so the feed is worth syndicating |
| AI search and new channels | Otterly.AI, Peec AI, AthenaHQ, Searchable | Track whether ChatGPT, Gemini and Perplexity recommend your products, with SKU-level shopping analytics on the mid-market tiers; roughly $29 to $600 a month as of September 2026 | Brands above about $2M where AI referrals are already showing up in GA4 |
| AI search and new channels | Profound, Evertune, Goodie | Enterprise AI visibility with prompt demand data, AI ad buying and coverage of Alexa for Shopping, Perplexity and Walmart surfaces | Brands above $10M, or omnichannel brands selling on Amazon and Walmart |
| AI search and new channels | ChatGPT ads | Self-serve sponsored placements, available across 31 European markets since August 2026 | Worth a small test budget in 2027 once you're measuring organic AI visibility; the inventory is new and CPMs haven't settled |
Three things to know before you sign anything. First, the AI visibility market is consolidating fast: in 2026 alone Adobe bought Semrush, Sitecore bought Scrunch, and a bunch of vendors quietly killed their self-serve tiers, so any price older than 90 days is a guess. Second, OpenAI pulled ChatGPT Instant Checkout in March 2026, so AI shopping answers now send buyers to your own site. That makes your product pages and feed quality the real lever, not some agentic checkout integration. Third, most shopping-visibility data is sampled from prompts and wobbles from run to run. Read the trend, not the single score.
Fix it in this order
The order matters more than the tools. Each step is cheap, and each one changes what the next step should be.
- Check the number is real. Compare platform CAC to what the bank account says. Run one free conversion-lift test on your biggest channel. If retargeting and branded search are claiming more than 20 to 30% of your 'new' revenue, cut them first and watch what the blended figure does.
- Split paid-new from paid-returning. Exclude existing customers from prospecting, sync the exclusion lists weekly, and report new-customer CAC on its own. For a lot of brands this alone explains most of the rise.
- Fix the site before the ads. If mobile conversion, page speed or checkout completion fail the diagnostic, nothing you do in Ads Manager will stick. A month of landing-page and offer tests usually moves CVR more than a quarter of media tinkering.
- Then feed the algorithm. Only once the site converts is it worth cranking up creative volume. Ten to twenty new concepts a month, tagged and tracked, with the losers killed inside a week.
- Then widen the funnel. Get AI-referred traffic showing in GA4, switch on catalogue syndication, and treat AI search as the channel you're learning for 2027, not the one that's going to rescue Q4.
Two things not to do. Don't cut spend across the board because a blended number went up; you'll trim the efficient channel and the inefficient one by the same amount and learn nothing. And don't buy a measurement platform to tell you what a spreadsheet and a holdout test would have told you in a week. The expensive tools are for brands that have already found the bottleneck and need to manage it at scale.
A useful tool starts with a clear brief. Explore the profiles below to find the questions worth asking.