Which action should count as a conversion?

Choose an action you would recognize as business progress. A conversion is the completion of a chosen goal, and that goal can include a purchase, subscription or contact submission. These examples appear in the CRO definition.

My recommendation is to write the goal before discussing design. For a service business, use a completed enquiry that meets your basic requirements. For an online shop, start with completed purchases and keep basket activity as a separate measure.

Hypothetical example: a local cleaning service

Imagine a cleaning company whose website offers both a downloadable checklist and a quote form. Choose quote requests as the main outcome for its service page. Record checklist downloads separately, then ask the sales team which enquiries match the area and work you actually cover.

Write a short acceptance rule: what makes an enquiry useful, who checks it and where the decision is recorded. Avoid changing that rule after seeing the results. If your goal is appointment booking, use the online booking system guide to work through the next step.

How do you calculate conversion rate consistently?

For a visitor-based rate, divide visitors who complete the goal by all relevant visitors, then multiply by 100. Keep the reporting period and page scope consistent. The underlying calculation is explained in this CRO overview.

Conversion rate = visitors completing the goal ÷ relevant visitors × 100.

The source illustrates this with 30 newsletter sign-ups from 200 visitors, giving 15%. Treat that as a worked calculation, not an industry benchmark or a promise for your website. See the original calculation example.

Write the counting rule above your report

• Name the action and the page being evaluated.
• Record the date range beside the numbers.
• Keep repeated actions separate from the count of people completing the goal.
• Label any session-based measure separately from a visitor-based measure.
• Keep completed orders and cancelled orders distinguishable.

I would also keep revenue, refunds and the quality of enquiries beside the main rate. Decide what would make an apparent improvement commercially disappointing. If you plan to change a discount, review the money left from each sale before approving the change.

What should you measure in Google Analytics 4?

Start with the event that represents your chosen outcome. In Google Analytics 4, or GA4, a key event measures an action particularly important to your business. Identify or create the event, then mark it as a key event. Google explains the setup principle here.

Give your website specialist a plain instruction: record a successful enquiry, not merely an attempt to send one. Ask them to demonstrate a successful submission, an unsuccessful submission and the resulting records. Compare those records with what reaches your inbox or order system.

Google documents a Key events column in reports including Landing pages, with the option to select a particular event. It also distinguishes Analytics key events from Google Ads conversions, which can be created from them. Read the reporting explanation.

Ask for a measurement handover

Request the event name, its trigger and the report you should open. Add the person responsible for checking it after website changes. For campaign links, use a consistent naming plan from the UTM tracking guide; UTM parameters are labels added to campaign links.

How can you find what stops visitors?

Inspect the journey before proposing a new layout. Conversion funnel analysis means examining the steps towards your chosen outcome and finding where people stop. User research and funnel analysis are both part of the approach described by Optimizely.

Choose one route, such as service page, quote form and confirmation. Follow it on your phone. Then ask someone unfamiliar with your business to complete the same task while explaining what they expect at each step.

Record observations separately from explanations. “They searched for delivery information” is an observation. “They do not trust us” is a hypothesis until you investigate it.

Use a customer journey map to connect the website task with what happens afterwards. Keep this first version brief: the customer's question, the action and the person responsible for the response.

Which landing page changes should you consider first?

Prioritize a missing answer or an awkward step you have actually observed. A landing page is the page a visitor reaches through a link or campaign. Clear calls to action, useful content, mobile usability and trust signals are among the areas covered in Optimizely's improvement checklist.

Hypothetical example: an independent furniture shop

Suppose visitors can see a table's photographs but cannot find its dimensions or delivery conditions. Propose adding those details beside the purchase decision. Record the idea as a hypothesis; do not present an improvement in sales as an established result.

• State what is offered and who it suits.
• Explain what the displayed price includes.
• Use a button label that names the next action.
• Review whether each form field is necessary at this stage.
• Show genuine reviews with relevant context.
• Complete the whole process on a mobile phone.

For the opening message, work through the value proposition guide. Write the customer benefit in ordinary language and check every promise against what your team can deliver.

For evidence, review the customer content guide. Choose material that answers a real buyer question. Avoid filling the page with decorative testimonials that your team cannot verify or explain.

How should a small business approach A/B testing?

Write a testable explanation before making a variant. A/B testing compares two alternatives; multivariate testing examines combinations of changes. Both are used to validate CRO hypotheses, as described by Optimizely.

Use a short experiment brief

For the hypothetical furniture shop, the brief might say: visitors ask about dimensions, so we will make dimensions visible beside the purchase button. The main measure will be completed purchases. The review will also consider cancellations and questions received by support.

For a first experiment, I recommend one focused change. Agree how visitors will be assigned to variants, what counts as success and when the test will end. Ask the person running it to document the required observation volume before starting.

Statistical significance helps assess whether a test result could be explained by chance. The CRO methodology overview discusses this principle. Do not declare a winner merely because one number currently looks larger.

If the evidence remains weak, record the result as inconclusive. For a site with little traffic, I would prioritize fixing demonstrable errors and observing customer tasks before commissioning a complicated experiment. Keep comparisons between different weeks labelled as observations rather than treating them as a controlled test.

Give any rollout an owner. Check the completed customer journey again after the change goes live, then confirm that measurement still works. Keep the previous version available to your website team in case the new experience needs correction.

What belongs on your first weekly CRO checklist?

Aim to finish the week with reliable measurement and one justified improvement to investigate. You do not need to promise a winning experiment within the same week.

1. Select: choose one page and its primary outcome.
2. Verify: complete the action and inspect its recorded event.
3. Observe: watch someone attempt the customer task.
4. Prioritize: choose the clearest obstacle supported by evidence.
5. Prepare: write the change, measurement rule and review conditions.
6. Assign: name the person who will make the decision.

Keep the decision log simple. Include the observation, proposed explanation, change and remaining uncertainty. Review it with the person handling customers, not only the person managing advertising.

If you are connecting this work to a broader acquisition plan, use the small business growth marketing guide. Keep the website experiment specific enough that the team can explain what it learned.