Conversion rate optimization

Conversion rate optimization, usually shortened to CRO, is the practice of increasing the share of visitors who complete a defined action: a purchase, a signup, a booking, a form. It combines research to find what stops people with experimentation to confirm that a proposed change actually helps.

How does conversion rate optimization work?

In four steps, and most programs are weakest at the first.

Find where people drop. Funnel analytics and page-level data show the steps that lose the most visitors relative to their traffic. Work out why. This is the diagnostic step, and it needs something analytics cannot supply: session-level observation, usability testing, or asking the visitor. Change something. A copy change, a form field removed, a reordered page, a clearer label. Prove it. Run the change against a control until the result is stable enough to keep.

Skipping step two is the most common failure pattern in the discipline. A team that knows where the drop is but not why generates hypotheses from the room rather than the customer, then tests them one at a time. That is an expensive way to guess, and it is why so many CRO programs produce a long list of inconclusive tests.

What counts as a conversion?

Whatever the business decided counts, which is a real problem in practice.

Macro conversions are the ones on the board report: revenue, qualified leads, completed applications. Micro conversions are the steps toward them: an account created, a size guide opened, a store locator used. Optimizing micro conversions is legitimate when they genuinely predict the macro one and misleading when they do not, and very few teams have checked which is true on their own site.

Define it too narrowly and you will optimize a page into a higher rate on a step nobody needed. Define it too broadly and every number goes up while revenue does not move.

What is a good conversion rate?

The comparison that matters is with your own site last quarter, not with a published benchmark.

Rates vary by an order of magnitude across industry, traffic source, device, product price and intent. A high-consideration purchase, a regulated financial product and a repeat grocery order have almost nothing to say to each other. Published benchmarks are averages across all of that, and the main thing they are useful for is starting an argument in a planning meeting.

The more useful framing is composition. A site-wide rate is the weighted average of very different audiences, and it can fall while every individual segment improves, simply because the traffic mix changed. Segment before you conclude.

What are the limits of conversion rate optimization?

The frame is narrower than most of what actually happens on a website.

CRO is defined by a conversion event, so work with no conversion attached falls outside it. Someone looking for a return policy, a branch location or an explanation of a coverage rule has a real task and no purchase intent. Support deflection in particular fits badly: a visitor who resolves their own question and leaves satisfied is a success that no conversion metric records. Teams whose only lens is CRO systematically underinvest in those journeys because their instrumentation cannot see them.

It also inherits the limits of experimentation. Tests need traffic, they need patience, and they answer which variant won rather than why. And because CRO sits closest to revenue, it attracts pressure to produce a win every quarter, which is how a discipline built on rigor ends up shipping changes that were never conclusive.

How Pulse Insights relates to this

We are used for conversion rate optimization, and the part we change is the diagnostic step that most programs skip.

Instead of inferring why people abandon a step, we ask them, in the moment, with a single short question triggered by the behavior itself. The answer routes to a response the client's team approved in advance, so the visitor who told you what was blocking them can get it addressed before they leave rather than joining a backlog. That is the difference between a research finding and a recovered session.

The mechanism is the same one that makes our surveys work: contextual placement produces response rates of 2-10x the industry standard, and placements at moments of resolved intent, such as confirmation pages, run far higher again.

One honest caveat, and it is why conversion rate optimization is our secondary category rather than our primary one. A good deal of what we do has no conversion event attached: wayfinding, product discovery, self-service resolution. If your problem is entirely a checkout funnel, CRO is the right frame. If it is broader than that, the CRO label will scope the project too tightly.

Frequently asked questions

What is conversion rate optimization?

The practice of increasing the share of visitors who complete a defined action, using research to identify what stops them and experimentation to confirm that a change helps.

What is the difference between CRO and A/B testing?

A/B testing is the verification method. CRO is the whole discipline, including the research that decides what is worth testing. A program that is only running tests has automated the cheap half.

How long does a CRO test need to run?

Long enough to reach a stable result and to cover at least one full business cycle, which for most sites means a minimum of two weeks. Stopping when the number looks good is the most common way to ship a change that does nothing.

Why do most CRO tests fail to produce a lift?

Because they test hypotheses generated internally rather than causes established with customers. The tests are usually run correctly; the ideas going into them were guesses.

Is CRO the same as experience optimization?

No. CRO is scoped to a conversion event. Experience optimization is broader, covering wayfinding, comprehension, product discovery and support outcomes that no conversion metric captures.

Related: Experience optimization · A/B testing · Digital friction · Pulse Insights vs Optimizely

Browse every term in the Pulse Insights glossary.