A week inside the CRO program of a 9-figure US supplement brand

How to read this week

This is the CRO program we run for a 9-figure US supplement brand that sells subscription-first through Shopify and spends heavily on Meta. The brand has its own media buying team; Succession runs the conversion layer: landing pages, offers, checkout and the daily read of what is working. I lead it, with Omar, our AI search lead, building the data pulls and the pages we test.

Every episode below happened. They come from July to October 2026 and are arranged as one working week so you can see the rhythm of the job. The brand is not named and no revenue or prices are shown: figures are rates, percentage changes and session counts. All numbers are US traffic only.

Monday: is conversion rate really up?

The week starts with a question: store-wide conversion rate looks unusually high. Everyone wants it to be the new pages. The first job is to check before anyone celebrates.

Step 1: clean the number

Store-wide conversion rate was not high at all once we looked properly. A store-locator page with no buy button had grown to about a quarter of all US sessions, which drags the store average down. With that page stripped out, the clean rate was the best in a month: up 0.40 points on the week before.

Step 2: find what actually moved

• The biggest driver was ad spend, not pages. Meta spend had been cut 28% from its peak two days earlier. Fewer, better-targeted clicks lift conversion rate mechanically. Cost per purchase improved 15%, but revenue sat about 9% under the peak day.
• The real page gain was small. On days with the same Meta traffic volume, before and after the new landers, conversion rate rose 0.13 to 0.30 points. Real, but not the headline.
• Volume decides more than design. On days with about 1.8 times the paid traffic, conversion rate ran 1.4 to 1.6% instead of 1.9 to 2.0%. Any conversion rate read that ignores volume is really measuring spend.

The best day that was not

Nine days later the same question came back in another form: is today one of our best days of the year? Counted with subscription renewals, the morning ranked 23rd of 365 by orders. Renewals bill at midnight and were at record levels, so they front-load every morning. Without them, the same morning ranked 179th. A normal good day.

What you can copy: before you report a conversion rate, strip pages that cannot sell, compare days with matched traffic volume, and split out subscription renewals. Our guide on why a Shopify conversion rate suddenly drops walks through the same checks in reverse.

Tuesday: media buying or CRO, which side moved revenue?

Every brand with an in-house media team and an outside CRO team eventually asks who is moving the number. Arguing about it is useless. Splitting it is easy.

Revenue per ad dollar is three numbers multiplied: sessions per ad dollar (the click side, owned by media buying) times conversion rate times average order value (the page and offer side). Take the logs and the three changes add up, so each side's contribution is visible.

The page side gained; the click side lost. But the honest read has three caveats, and they matter more than the verdict:

• The media team did something hard. CPM stayed flat while spend nearly doubled. Frequency rising 32% is real audience saturation, not bad buying.
• Order value was mostly an offer decision. It rose because discounting fell and buyers shifted to the bundle. Strip it out entirely and the page side still wins: +5.4% conversion rate against -12.8% sessions per dollar.
• The last dollars did not pay. From August to September, daily spend went up and daily revenue went slightly down. That is a scaling problem, not a team problem.

The irony we keep repeating: conversion rate is the CRO scoreboard, but traffic volume moves it more than any page change. That is why we put both sides on the same table.

What you can copy: put revenue per ad dollar, sessions per dollar, conversion rate and order value on one sheet, by week. Our guides on Meta ROAS vs Shopify revenue and where your next $10K should go build the same sheet.

Wednesday: hundreds of sessions, zero add-to-carts

A new advertorial page went live. By morning it showed 403 sessions, zero add-to-carts and 93% bounce. The same template converts well on a sister page, and our end-to-end test (button, cart, checkout) passed twice. So the page was not broken. The traffic was not real.

The signature of Meta's review crawler

• Sessions arrived in one-hour bursts that matched the times ads were created or edited, not when they delivered. One burst of 192 sessions lined up with a batch of new ads; another of 174 lined up with 86 ads being edited to add the page as a site link.
• Most hits carried a right-column placement tag. Real right-column delivery across those ads, lifetime: $6.17 and 3 clicks.
• Heavy Sweden and Ireland share (where Meta runs data centers), desktop or unknown devices, 97 to 100% bounce, no UTM tags.
• The proof: the ads that used the page as their main link were paused with $0 spend and zero impressions, ever. They still produced 226 sessions.

Of 518 sessions in the page's first day, 29 came from real US placements. Those 29 added to cart at 10%, about the same as the control page at 9.1%. The page was fine all along.

The same crawler later polluted an A/B test: about 5,000 phantom visitors in two hours after a batch of ads was created, which the testing app counted and Shopify mostly filtered. Those windows get cut out before any test is read.

What you can copy: before calling a new page broken, check that an ad using it as its main link has impressions. At a 1.5% conversion rate, zero orders only means something after about 200 real sessions. The full method is in why a new landing page gets traffic but zero add-to-carts.

Thursday: can we hit 5%, and should we skip the cart?

Thursday's question was ambitious: can the main landers reach a 5% conversion rate? The honest answer had to come before the plan.

The ceiling

In 40 days there were 93 page-days where a lander had at least 3,000 US sessions. Not one reached 3%. The all-time best was 2.38%. And design was not the limit: four pages on two templates, with every new section, gallery and countdown we shipped, all sat between 1.52% and 1.71%. Page changes moved the number by about 0.2 points either way.

The leak was checkout. On the advertorial lander, 8.45% of visitors reached checkout but only 20.3% of them paid. The offer is subscription-only, the traffic is 100% cold mobile social, and that combination pushes many undecided people into checkout.

The test: skip the cart

We cloned the bundle lander one to one. In the test version, the main button went straight to checkout instead of opening the cart drawer. Split 50/50, judged on conversion rate per visitor only. The direct-to-checkout version converted at 2.46% against 2.24%, with about 9% more revenue per visitor. We rolled it out to the live page the same week.

Then we walked our own checkout like a buyer and compared it with ten big supplement brands (Seed, AG1, Ritual, Obvi, Pendulum, Grüns, Arrae, Happy Mammoth, Cymbiotika, Bloom). Things we found on ours: a "hurry" timer counting up from zero, an upsell above the email field, free gifts listed as bare "FREE" lines with no value shown, and a delivery promise that said 1 to 2 days on the page but 3 to 5 business days at checkout. Two of them, AG1 and Bloom, price their free gifts and zero them with an automatic discount, so the checkout shows a crossed-out price and total savings.

The realistic target we gave: 3.0 to 3.5% on the main lander, which would be a store record, by fixing checkout first, not by adding page sections. What you can copy: read your funnel step by step before redesigning anything. Our guide on add-to-carts with no purchases shows how to find the leaking step.

Friday: the result that shows up a month later

The work that paid shows up a month later, at renewal. Because the landers are subscription-first, every new cohort turns into renewals a month later. Subscription renewals had their record week, renewal revenue for September came in at more than double August's, and renewals went from 11% to 34% of the store's net sales in seven weeks, with no ad spend behind them.

Honest footnote: new subscribers grew 48% like for like, but mostly because ad spend rose 37% with cost per new subscriber flat. And renewal refunds rose to 12.6% of gross, up from 7 to 8% in August. That is the next thing we are watching.

What you can copy: judge a subscription lander on the renewals it creates, not only the first order. Our guides on cutting subscription churn and LTV to CAC for subscription brands cover the other half.

What the week adds up to

None of these days had a redesign in it. The work was: a clean number every morning, the same table for media and CRO, a crawler filter before any read, and one change at a time with a fixed read window. The page work that did happen, a new lander, a skip-the-cart button, a checkout teardown, came out of those reads.

At Succession this is the routine on every account: one market, matched volume, bots filtered, one scoreboard agreed before anything ships, and changes shipped the same day. You can see the longer result for this brand in our case study on lifting paid conversion rate 35% at matched traffic.

1. Strip pages that cannot sell from your conversion rate and say which number you report.
2. Compare days with matched traffic volume before calling any lift.
3. Put sessions per ad dollar, conversion rate and order value on one weekly sheet.
4. Filter Meta's review-crawler sessions before judging a new page or an A/B test.
5. Read the funnel step by step and fix the leaking step before adding page sections.

FAQ