Joe Smith — available this week, Eastern time, 30–40 hours a month.

Five years in-house as the traffic and conversion owner for a consumer brand, until the company sold. I ran the paid traffic, almost all Google and Meta, and the product messaging it landed on. Acquisition cost was over half of revenue. The biggest win wasn't a button color: we repositioned each main product around who it was for ("if you're X, this is for you; if you're Y, here's the one for you"). Acquisition cost fell to 18% of revenue.

Using Test 1 below (hold the pop-up): I build the A/B activity in Target myself — audience, the two experiences, and the goal metric (Amazon clicks per visit, with Attribution sales as the second check). Before launch I QA every experience on desktop and phone with Target's QA links and confirm the goal fires in the analytics. A developer is needed only if the pop-up is hard-coded in the theme. Then it runs to a sample size set in advance, never stopped early on a good day, and the readout goes out in plain words.

I start with the scoreboard, not the pages: is every lead counted once, tied to its source, and does a "lead" on the site match a lead in sales? Then I read the last six months of tests. If tracking is sound and tests ran to full sample, I refine. If not, I fix measurement first. Month one: about 10 hours on the audit, 10 on a ranked roadmap of the 2–3 pages with the most traffic and biggest drop-off, and the rest launching the first test.

A store owner's Meta account showed about 40% more purchases than his store had. What I told him: "Facebook is counting some of your sales twice, so the ads look better than they are." The cause was an old pixel left over from a migration. I recommended removing it before touching any budget. He worried the drop would look like the ads got worse overnight, so I showed both counts side by side for a week first. After the fix, Meta matched the store.

The Amazon links already carry Amazon Attribution tags (maas), so each test is scored on real Amazon sales, not just clicks. A version with more clicks and fewer sales loses.

For a quote-request or service-call form, the first three tests I'd look at:

1. Ask for the zip code first, alone. One easy box starts the form; people who start usually finish. Scored on completed leads, not form starts.
2. Show the price range or "from $X" before the form. Fewer leads that hang up at the quote, more that book. Scored on booked jobs from sales, not raw leads.
3. Put the phone number in the sticky header on phones. Many home-services buyers would rather call. Scored on calls plus forms together, so one doesn't just steal from the other.

Every one is judged on what sales counts as a lead, not on what the form counts.