What a Conversion Rate Optimization Agency Should Fix Before Running Tests

A conversion rate optimization agency should not start with a test calendar. If the offer is unclear, the tracking is broken, and the page has three competing actions, you are paying to randomize noise.

I say this as someone who designs and builds the pages those tests run on. The unglamorous work is what makes a later experiment meaningful.

What 'CRO' is supposed to mean

CRO is a loop: understand behavior, change one meaningful thing, measure, keep or revert. Tools are optional. Honesty about the offer is not. If the company cannot state the action they want and the audience they want it from, there is nothing to optimize.

Fix these before anyone talks about variants

1. Tracking that matches reality

If demo requests, booked calls, and qualified pipeline are not events you trust, stop. You will pick a winner that increased clicks on a button that does not create revenue. Confirm the same conversion definition in ads, analytics, and CRM.

2. The offer

Write the promise in one sentence. If two teams write two sentences, you do not have an offer. You have a compromise paragraph. Tests cannot resolve a political argument.

3. One primary action

A page with Book, Start trial, Download, and Chat is not a test candidate. It is a buffet. Pick the action that matches the traffic source.

4. Proof next to the risk

Move one real proof element next to the claim that scares people: price, timeline, or 'does this work for companies like mine.' Footer logos do not do that job.

5. Mobile completion

Complete the funnel on a phone. If you cannot book or submit in under a minute, you do not need a headline test. You need a path test, which is often just deleting steps.

A/B testing a broken booking flow is how agencies stay busy and companies stay confused.

What a good CRO partner does in month one

1. Rebuild the conversion taxonomy.
2. Watch recordings and list the first five friction points.
3. Kill or hide the extra CTAs.
4. Ship one structural change, not ten micro copy tweaks.
5. Only then open an experiment tool.

What a weak CRO pitch sounds like

• We will run 12 tests in 90 days.
• We guarantee a lift percentage.
• We need access to everything on day one and a redesign budget on day two.
• We do not need to understand the sales process.

Guaranteed lifts are a tell. Traffic mix changes. Seasonality exists. Anyone who promises a number without a baseline is selling a story.

Tests that are worth running later

After the structure is honest, test things that change the decision: offer framing, social proof type, pricing presentation, form length, and the first section's job. Do not start with button color unless you have already won the boring war and have enough traffic to detect a tiny effect.

For the design layer of this work, I use the same rules as conversion-focused web design. CRO without that foundation is a theater subscription.

How to brief an agency without wasting a quarter

Give them the money pages, the conversion definition, the sales notes on why deals die, and access to recordings. Ask them to return a 'do not test yet' list. Hire the team that is willing to write that list.

If you want that list from someone who also has to build the page afterward, I do that work. Send the URL and the event you call a conversion.