A Framework for Testing a New Vertical
A structured way to expand into a new vertical without risking your best-performing traffic.
Expanding into a new vertical is one of the more reliable ways to grow earnings, but testing it carelessly can also tank your account-level metrics. Here's a structured way to approach it.
Start with a capped test budget or traffic share
Rather than redirecting a large share of traffic into an unproven vertical, carve out a small, fixed percentage — commonly 5-10% — so a weak early result doesn't distort your overall performance while you're still learning.
Pick a comparison baseline
Before launching the test, write down your current best-performing vertical's EPC. That's the number the new vertical needs to approach or beat to be worth scaling — not an abstract "good" number, but your own actual baseline.
Give it a defined window, not a vibe check
Two to three weeks is usually enough to get a meaningful read, assuming your traffic volume is high enough to produce statistically useful conversion numbers. Cutting a test after two days of noisy data (or letting it run for months out of inertia) both lead to bad decisions.
Decide your scale-or-cut criteria in advance
For example: if a two-week test in a new vertical lands EPC within 20% of your baseline, scale traffic gradually; if it's well below that after a fair test window, cut it and try a different offer within the vertical before abandoning the vertical entirely — sometimes it's the specific offer, not the category, that underperforms.