How Landing-Page CR Lifts Compound With Meta Algorithm Learning
Higher post-click conversion rate feeds Meta's auction a stronger signal, dropping CPM and CPA in parallel. The result: CAC falls more than the raw CR lift — and compounds over 4-8 weeks.
Quick answer
On Meta-heavy accounts, a 15% landing-page CR lift typically drops CAC by 20-25%, not 15%. Higher post-click conversion rate feeds the auction a stronger conversion signal, which lowers CPM and CPA in parallel. The compounding curve takes 4-8 weeks to fully express, and gains plateau once the algorithm re-stabilises on the new conversion baseline.
How Landing-Page CR Lifts Compound With Meta Algorithm Learning
The mechanism by which a post-click CR improvement triggers a larger-than-proportional CAC drop on Meta as the auction re-learns on the new signal.
When you lift landing-page conversion rate, Meta's auction sees more conversion events per 1,000 impressions on your pixel. The bidding system reads that as 'this advertiser is converting better than its CPM band' and responds two ways: it lowers your effective CPM (because predicted value rose) and tightens delivery toward higher-intent users (lowering CPA). The two effects multiply, so a 15% CR lift typically delivers a 20-25% CAC drop after the learning phase re-stabilises. The effect is strongest on accounts where Meta is 50%+ of paid spend and where conversion volume per ad set clears Meta's learning threshold (~50 events/week).
Most teams model a CR lift linearly: 15% more conversions at the same CPC equals 15% lower CAC. On Meta-heavy accounts, that's wrong by roughly a third — and the gap is the auction learning loop.
Why the CAC drop exceeds the CR lift
Meta's auction prices each impression on predicted action rate × bid. When your landing page converts better, predicted action rate rises across every audience segment Meta is testing. That single input change reprices the auction in your favour.
Two things move at once. CPM falls because Meta will pay more (internally) to put your ad in front of users it now believes will convert. CPA falls a second time because the campaign exits learning faster and delivery concentrates on higher-intent placements and lookalikes.
The rough math
If LP CR rises 15%, expect CPM to drop 5-8% and CPA (at constant CPM) to drop another 13-18%. Stacked, that's a 18-25% CAC drop versus the pre-test baseline. The exact split depends on how much room your account has on the CPM side — accounts already winning quality auctions see more of the gain on CPA.
How to detect compounding in Ads Manager
The signal shows up in three windows. In week 1 you see the raw CR lift in GA4 and Shopify — CAC barely moves because Meta hasn't re-learned yet. In weeks 2-4 CPM starts drifting down on the winning ad sets while CPA accelerates faster than the LP CR alone would predict.
Watch the ratio of CPA delta to LP CR delta. If it's stuck near 1.0x after 14 days, the auction hasn't re-learned (often because conversion volume per ad set is under 50/week). If it climbs past 1.3x, compounding is active and you should expect another 5-10% improvement before plateau.
Typical 8-week curve after rolling out a 15% LP CR lift on a Meta-heavy DTC apparel account
| Week | LP CR vs baseline | CPM vs baseline | CPA vs baseline | CAC vs baseline |
|---|---|---|---|---|
| Week 1 | +15% | 0% | -12% | -12% |
| Week 2 | +15% | -3% | -15% | -17% |
| Week 3 | +15% | -5% | -17% | -21% |
| Week 4 | +15% | -6% | -18% | -23% |
| Week 6 | +15% | -7% | -19% | -24% |
| Week 8 | +15% | -7% | -19% | -25% |
| Week 10 | +15% | -7% | -19% | -25% |
How to capture the full curve (and where it plateaus)
Three operational rules protect the compounding effect. Don't restart ad sets — re-uploading creatives resets learning and wipes the CPM gain you've been earning. Don't widen budgets more than 20% per week during the learning re-phase; Meta treats sharp budget jumps as a new state.
The plateau hits around weeks 6-8 because Meta's predicted action rate model converges on the new conversion baseline. After that, additional CAC gains have to come from a new lever — fresh creative, a second LP win, or a margin-side change. This is why teams pair LP CR tests with creative-fatigue tracking; otherwise you can't tell which lever caused the next CAC move.
Don't re-attribute the plateau as test failure
A common mistake: at week 9 CAC stops improving, and the team concludes the LP test 'wore off'. It didn't — the compounding finished. CAC is now stable 20-25% below baseline, which is exactly the win you tested for. Re-baseline your CAC target and move to the next lever.
Experiment ideas to stack on top
Once the curve plateaus, the highest-ROI next tests are the ones that feed Meta a second conversion-signal upgrade. Adding an above-the-fold trust band on a beauty SKU PDP often delivers another 4-7% LP CR — small in isolation, but it triggers a fresh 6-10% CAC drop because the auction re-learns on the stronger pixel signal.
Sequencing matters. Run one LP test at a time, give it the full 4-8 weeks, and only then ship the next. Stacking two tests in week 2 makes attribution impossible — you won't know whether the CAC drop came from LP variant A, variant B, or the auction learning on either. Also worth pairing with an iOS signal recovery setup, since the compounding effect is stronger when Meta has cleaner conversion data to learn from.
Frequently asked questions
Because two things drop together. Meta's auction lowers your CPM when predicted action rate rises, and CPA drops a second time as delivery concentrates on higher-intent users. The two effects multiply, so a 15% CR lift compounds to a 20-25% CAC reduction once learning re-stabilises.
Typically 4-8 weeks on accounts that clear Meta's learning threshold (~50 conversions per ad set per week). Week 1 shows mostly the raw CR lift; weeks 2-4 add the CPM drop; weeks 5-8 squeeze the last 3-5% as the auction model converges on the new baseline.
Partially. Google's Smart Bidding also re-learns on a stronger conversion signal, but the CPM-side effect is weaker because search auctions are intent-priced rather than predicted-action-rate-priced. Expect a 16-19% CAC drop on Google for the same 15% LP CR lift, versus 20-25% on Meta.
Usually one of three things: conversion volume per ad set is under 50/week (Meta never exits learning), you restarted ad sets or changed budgets mid-test, or Meta is under 40% of paid spend so the blended CAC math dilutes the gain. Audit those three before concluding the effect isn't there.
It dampens it. With weaker pixel signal, Meta needs more conversion events to re-learn, so the curve takes 6-10 weeks instead of 4-8 and the final CAC drop is closer to 17-20% than 25%. Server-side CAPI and a clean Conversions API setup restore most of the compounding.
Cautiously — no more than 20% per week. Sharp budget jumps reset Meta's confidence in the predicted action rate and you forfeit part of the CPM gain. Once the curve plateaus around week 6-8, you can scale more aggressively against the new CAC.
Hold creative constant during the LP test window. If you can't, isolate by ad set: keep one ad set on the old creative + new LP, and one on new creative + old LP. The CAC delta on the first ad set is your clean LP-lift signal.
Less. Retargeting audiences are already high-intent, so the CPM repricing has less headroom. Expect retargeting CAC to drop roughly in line with the raw CR lift (~15%), while prospecting drops 20-25%. The compounding lives in cold-audience delivery.
No, assuming the LP variant stays live and creative doesn't fatigue. CAC stabilises 20-25% below the pre-test baseline. Drift upward later is usually creative fatigue or audience saturation, not the LP lift reverting.
Roughly 8% or more. Below that, the CPM repricing is too small for Meta's auction to react meaningfully, and you'll see a near-linear CAC response. At 10-15%+ the compounding kicks in reliably; above 25% the curve looks similar to 20% — Meta caps how aggressively it reprices any single advertiser.
Track CAC, channels, and funnel conversion in one place
Metricuno connects ad spend, funnel events, and revenue so you can see CAC by channel, cohort, and campaign — without stitching together five tools.