How Smart Funnels adds a second layer of wins underneath the page tests you already run through Meta CBO.
Whichever page is taking more spend at a better CPA is winning. It is fast, but it can only detect big swings.
One number is what the test reports. Flip it to see what the same page looks like change by change.
Your CBO test is picking the page. Smart Funnels then keeps improving that page, element by element, and moves traffic to the better versions while they’re running.
We import the page and build 10 to 30 versions that each change one factor: headline, hero image, subtitle, benefit block, testimonial, button text, price display, etc. Your page stays your page.
Every click from the same ad is routed to one of the versions the moment it lands. One ad, one audience, thirty pages, all at once.
Over time the algorithm learns which page or element combination converts for what type of users. Smart Funnels can maximize uplift by sending each user to the page that is the most likely to get conversions for each individual.
Big swings stay in CBO. Whenever a new page wins there, it becomes the next page we iterate on. The two layers stack: CBO finds the next page, Smart Funnels compounds it.
Every figure below is from a live test on paid traffic.
Same account, same week, same product. The client’s ad set pointing at the Smart Funnels winners bought purchases at $84.68 while the client’s own CBO testing campaign paid $157.85.
73 funnel variations tested across two rounds on an advertorial-to-product-page funnel. The winning variation changed two elements on the product page: a subtitle under the product title and a rewritten benefit block. The next iteration is already designed around what won.
Two nearly identical before-and-after collages. Same page, same ad, same offer; only the top image differs. Version B earned 55% more revenue per session than the test average, Version A earned 48% less. Measured 10-02-2026 to 26-02-2026 on live, model-routed traffic; one session is one ad click. That difference is invisible to a CBO test and it sits on one element.
What to test next is now obvious: the winner reads before-to-after left to right, the loser reads right to left (a Hebrew-market page), and the winner shows more faces uncovered. Two hypotheses, each a single element, each a small win that stays.
Four top images on one page, 36 page versions. The younger couple without the animated overlay came first in both reads of the test, at +37% against the test average. The older couple with the overlay came last, at −54%. Now these findings feed the next image round.
Big swings are how you find a new best page. They are a poor way to keep improving it: every swing changes too much at once to learn from, and the better the page already is, the more often the next swing lands below it.
A big swing changes the headline, the image, the layout, the offer and the copy at once. Some of those changes are lifting CVR, others are pulling it down. Meta shows you the net. The winner is carrying losers you never see, and the pages that lost were carrying wins you threw away.
Going from 1% to 2.5% on one lucky swing is what makes CBO testing feel great. From there, most big swings land below the winner. Test twenty pages in a year and one or two beat the best page; the rest lose, and every losing test is spend at a worse CPA.
Because nothing is isolated, nothing carries over. You do not learn why a page lost any more than why one won, so every big swing starts from zero. Smart Funnels starts from the winner and changes one factor (which can be multiple elements!) at a time, so each round’s answer feeds the next.
Your numbers. Put in what your next big swing does, then see what steady element wins stacked on top of it look like over twelve months.
Element wins compound: each round’s lift applies to the page as it is at that point, including every big swing. Red dots are the big-swing tests that did not beat your best page. Edit any input.
No. The money does not wait. Significance is a separate question, and whether you want to answer it is up to you.
The algorithm moves clicks to the versions that convert better in real time, the same way CBO moves spend between pages. The CPA and ROAS you watch in Meta start moving from the first days of the test.
Waiting for 95% is what it takes to say which element caused the lift, so the next round builds on a known lever instead of a guess. It is never a gate on the profit.
Learn: keep more versions in play long enough to name the winning element. Exploit: push nearly all traffic to whatever converts and take the profit now. Flip it on the right.
Your CBO tests keep running. Ours runs on the page that wins there. If a new page wins in CBO next month, it becomes the next page we iterate.
We design the element variations from your winning page and agree the list with you.
Every version built, checked on desktop and mobile, tracking verified.
Test live. Traffic starts shifting to the better versions the same day.
Winners roll into the page. Next round starts from what won.
Pick the page that is winning in CBO right now. In about a week it is running thirty versions of itself on the same traffic.