What is conversion rate optimisation, and where should you start?

Conversion rate optimisation (CRO) is the process of improving the proportion of visitors who complete a useful action, such as buying or becoming a qualified lead. You identify obstacles in their journey, make evidence-based changes and evaluate the result. The goal is more valuable business from the traffic you already receive.

Start with one journey and one commercial outcome. A higher form completion rate is little help if the additional leads never qualify.

How do you calculate conversion rate?

Divide the number of eligible visits or people that converted by the total eligible visits or people, then multiply by 100. State which denominator you use, what counts as a conversion and the reporting period. Comparing a user-based rate with a session-based rate can give a misleading impression of performance.

Consider a hypothetical furniture retailer. In one month, 10,000 eligible sessions include 200 sessions with a completed purchase:

Session conversion rate = 200 ÷ 10,000 × 100 = 2%.

These figures are illustrative, not a client result. Count converting sessions once, even when a session contains multiple orders.

For lead generation, record both form completion and subsequent qualification. Removing a budget question might increase submissions while leaving the number of suitable sales conversations unchanged.

Where should you look first?

Choose a valuable journey you can measure, such as product page to purchase. Before interpreting drop-off, complete the journey yourself on the affected device and check that recorded actions match what happened. A missing purchase event needs a measurement fix; a payment failure needs a technical fix.

Google's GA4 funnel exploration documentation explains that a closed funnel requires entry at the first step. Users who skip a required step are excluded from subsequent steps. A missing cart event can therefore make a successful shopper disappear from your funnel.

For our hypothetical retailer, a consistently defined, closed user funnel might show:

These hypothetical user counts describe a specific journey. They use a different denominator from the sitewide session rate above.

The largest percentage drop is not automatically the biggest opportunity. Many product viewers are browsing. Compare the same stage across devices, traffic sources and comparable periods before selecting a problem.

Suppose the retailer's mobile cart-addition rate fell while desktop remained stable. Check whether mobile traffic composition changed, then inspect the product journey on mobile. That finding narrows the investigation; it does not prove that the layout caused the decline.

Our product views without cart additions guide covers that diagnosis in more depth.

Turn a suspected problem into evidence

Investigate one question: what prevents suitable mobile shoppers from deciding whether this furniture fits their space?

Use evidence that can answer it:

• Support enquiries: Look for questions about dimensions, assembly or delivery. Record the relevant product and the information customers could not find.
• Observed sessions: Look for repeated attempts to open specifications or interact with an unresponsive element. Behaviour suggests a question; it rarely establishes motivation.
• Usability tasks: Ask representative shoppers to choose an item for a space with given dimensions. Observe whether they can make the decision.

Nielsen Norman Group's task-scenario guidance recommends realistic activities that avoid revealing how to use the interface. Telling participants to open the dimensions tab would conceal whether they could find it themselves.

In this hypothetical investigation, suppose participants repeatedly miss dimensions hidden inside a collapsed mobile section, and support enquiries raise the same question. That supports making dimensions easier to find. It provides no particular reason to change the purchase button's colour.

Record the observation separately from the proposal: “Participants could not find the dimensions” is a finding. “Show a dimensions illustration beside the product images” is a possible solution.

Should you fix it or run an A/B test?

CRO includes technical repairs, clearer information and experiments. Choose the evaluation method according to what remains uncertain.

Nielsen Norman Group's A/B testing introduction describes comparing variations against predetermined success metrics. For the furniture example, test the dimensions illustration against the current presentation if there is enough traffic to answer the commercial question.

Before launch, estimate the sample required using the baseline rate, the smallest improvement worth acting on, and your chosen statistical assumptions. Divide that requirement by eligible traffic to assess feasibility. A universal rule such as “run every test for two weeks” cannot account for those differences.

Write down what would change your decision

For the hypothetical furniture retailer, the test brief could be:

• Audience: Eligible mobile visitors to the affected product pages.
• Evidence: Repeated difficulty locating dimensions during usability tasks, supported by relevant customer enquiries.
• Change: Place a readable dimensions illustration beside the product images.
• Primary outcome: Purchase conversion among eligible users, measured consistently across variants.
• Supporting measure: Cart additions, to understand whether more shoppers progress.
• Commercial checks: Contribution per eligible user and size-related returns, allowing time for returns to arrive.

Specify the smallest worthwhile improvement and acceptable commercial limits before launch. Keep assignment consistent and follow the planned analysis method.

If cart additions rise but purchases do not, the change has not demonstrated the intended sales benefit. If the purchase estimate remains too uncertain to distinguish a worthwhile gain from harm, report an inconclusive result. An adequately precise estimate that excludes the intended gain weakens the hypothesis.

Your first CRO task

ConversionNest's view: start with a problem you can describe and investigate before commissioning a redesign.

Choose one valuable journey, write down its conversion definition and check the measurement. Identify one stage worth investigating, collect evidence about the obstacle, then document the proposed change and evaluation method.

Keep that record after the change ships. A before-and-after increase alone cannot isolate its effect from seasonality, traffic changes or concurrent work. Your first useful output is a defensible decision about what to fix, test or investigate next.