Offer Tests Sized to a Decision, Verticals Scouted

Sixty conversions on the leading arm. Below that the interval around earnings per click is wider than the difference between the two offers you are choosing between, and the decision you make is a coin toss you paid for. Every test I design is sized backwards from a decision someone will actually take — go, drop, or renegotiate the payout — and if the budget does not reach sixty on the arm we expect to win, I will tell you the test is not worth running rather than run a smaller one and write it up confidently. How a vertical gets shortlisted: Demand evidence before offer evidence. For app verticals that means install and revenue trajectories in Sensor Tower or AppMagic read across four quarters, not a single month, plus the shape of the subscription price ladder in the top ten apps — a category where everyone has converged on the same annual price is a category with no room for an affiliate margin. For apparel it is returns rates and resale velocity, because a 38 percent return rate eats the commission on the sale that generated it. For donor programmes it is the second-gift rate the organisation can actually report. I shortlist three verticals, and I write the argument against each one as well as for it. How the offer itself gets checked: The payout is the least interesting number on the page. I ask the advertiser for their scrub rate in writing and their hold period, then I ask two current partners what they actually received against what was reported. I check the chargeback window against the payment terms, because an offer paying net-60 on a product with a 90-day refund policy has moved its risk onto you quietly. Cap size, day-part caps and the advertiser's history of pausing mid-month go into the same note. Roughly a third of the offers that look best on a network's top-offers page fail one of these before a single click is bought. What lands on your desk: A written recommendation of one to three offers with the test design attached: sample size per arm, the traffic sources the test is valid for, the stopping rule, and the number at which I would walk away. Then the read itself — EPC, conversion rate and post-hold revenue with the interval shown, not a single point estimate — and a plain recommendation. Where the answer is that nothing here clears your cost of traffic, that is the answer, and it has been the answer twice in the last two years. Not included: Media spend, which you pay directly. Introductions as a service — I will name the networks and advertisers worth a conversation and I will sit in the call, but I do not sell a contact list and I do not take a rebate from a network for placing you, which is why the fee is what it is. I also do not scout gambling, adult, nutraceutical or short-term credit verticals, and I will not review an offer in those categories as a favour. Who should not hire me: Anyone who wants a vertical named this week. The shortlist takes three to four weeks because the demand evidence takes that long to read properly. And anyone whose real question is which offer is converting for other people right now — that question has a market price of nothing, because by the time it reaches a listing the answer has expired.