How to use Experiment Watch: learn from competitor A/B tests you were never shown

Every funded competitor in your category is running conversion experiments right now, and you are seeing exactly one variant of each: the one their testing tool decided to show your IP address. The rest of the experiment, the alternative headlines, the reshuffled pricing tables, the variants that quietly lost, happens in a part of the market you were never shown. That invisible experimentation is some of the most expensive research in your category, and it is being paid for by someone else.

The Experiment Watch report makes that layer visible in two ways. First, it detects which A/B testing tools each competitor is running, which tells you who is experimenting at all and how seriously. Second, through weekly scraping from multiple IPs, it tracks what changed on their key pages and what stayed live over time, so you can see when a page is in flux and which version eventually won the right to stay. The result is evidence for your own funnel and messaging tests, which is why the report belongs at the start of any homepage, pricing-page, or signup-flow experiment plan.

Step 1: read the tooling layer as a seriousness signal

Start with who runs what. A B2B SaaS competitor carrying a dedicated experimentation platform is systematically testing; their page changes deserve the "observed experiment" reading. A competitor with no testing tooling whose homepage still changes weekly is redesigning on instinct, and their churn of copy tells you much less. This one distinction stops you from treating every competitor page change as a validated learning.

Step 2: read persistence, because survival is the verdict

The most useful signal in the weekly history is not what changed but what stopped changing. When an AI tooling competitor's hero cycles through three framings over six weeks and then settles on one for the next two months, the settled version is plausibly the winner of a real test with real traffic. Variants that appear and vanish within a week or two are the losers, and losers are valuable too: they are messaging directions a funded team paid to invalidate. Build your shortlist from both lists: settled patterns worth borrowing, and abandoned patterns worth avoiding.

Step 3: turn one observation into one reversible test (30 days)

• Pick exactly one observed experiment that maps to your funnel. If two developer tools competitors both settled on docs-first CTAs over demo-first, that convergence is a stronger candidate than any single change.
• Define the success metric before launch: signup conversion, pricing-page click-through, activation rate. One metric, measured the same way for the full 30 days.
• Define the stop condition with the same discipline: the number and the date that end the test either way. A reversible test with a stop condition is research; the same change without one is just a redesign you will rationalize.

The decision this report should produce is a single sentence in your experiment log: which observed competitor experiment we are testing in our funnel this month, what counts as success, and when we stop. Not five tests, one, because the point of borrowing the market's experimentation is to spend your own traffic on the highest-evidence candidate.

What Experiment Watch will NOT tell you

• The competitor's conversion numbers. No amount of external observation reveals the metrics inside their testing tool.
• Experiments running behind the login or inside the product. The report watches public pages; onboarding and in-app tests are invisible to it.
• Whether a winning pattern will win for you. Their traffic is not your traffic; the report supplies prioritized hypotheses, and your funnel supplies the verdict.

Before you plan the next homepage or pricing-page test, look at what the market already paid to learn. Run a free analysis at top-founders.com/analyze, open Experiment Watch, and let your competitors' testing budget shorten your own path to a converting page.