Conversion Rate Optimisation: Find the Real Bottleneck

Conversion rate optimisation starts with a more useful question: what is preventing the right visitor from completing the right action? The answer could be uncertainty about the offer, missing proof, an unexpected delivery charge, a broken mobile field or a request that arrives before the visitor is ready.

The challenge is to separate those causes before committing time and money to changes. A page with unsuitable traffic needs a different response from a page with qualified visitors and a failing form.

This guide explains how to locate the bottleneck, estimate its commercial significance and choose changes worth testing. It includes a practical diagnostic table and a worked example showing why a higher headline conversion rate can still produce fewer useful leads.

Define the conversion that actually matters

A conversion is a completed action, but actions have different values. Reading a pricing page, downloading a checklist and submitting a sales enquiry should not automatically count as equivalent success.

Choose a primary outcome for the page. For an online store, that may be a completed purchase. For a service business, it may be an enquiry that meets agreed qualification criteria.

Then track supporting actions to explain behaviour. Form starts, delivery information views and appointment selections can help diagnose friction. They should support the main outcome rather than replace it.

Use a consistent denominator

State whether your conversion rate uses sessions, users, clicks or another unit. Keep that definition consistent when comparing periods or variants.

For example, if you use sessions:

Session conversion rate = sessions containing the target action ÷ eligible sessions × 100.

Do not quietly switch to total events if one visitor can trigger the event several times. That changes what the number means.

Establish where the drop-off occurs

Map the smallest useful sequence between arrival and the desired result. Avoid building an elaborate diagram before checking the obvious breaks.

For a service page, the sequence could be arrival, form start, valid submission and qualified enquiry. For a store, it might be product view, cart, checkout and purchase.

Segment the numbers by device, source and landing page. A mobile problem can disappear inside a healthy desktop average. A campaign targeting researchers can make a good sales page look weak.

Observed pattern Question to investigate Visitors leave before engaging with the offerDoes the page match the promise that brought them? Visitors read but rarely begin the formIs the next step clear, credible and appropriate? Many form starts, few valid submissions Are fields, errors or requirements creating friction? Many submissions, few qualified leadsIs the offer attracting the right buyer?Strong cart activity, weak checkout completionAre cost, delivery, payment or technical issues appearing late?

The table suggests investigations, not automatic diagnoses. Confirm the cause with actual behaviour and direct testing.

Collect evidence from the people using the page

Analytics shows where activity changes. Customer conversations can help explain why.

Ask sales staff which questions appear repeatedly before someone commits. Review support requests, site-search terms and reasons prospects decline. These can expose information the page should have provided earlier.

Use usability sessions to observe a realistic task. Ask someone to find a suitable product, compare service options or request a quote. Watch where they hesitate without guiding them through the interface.

Session recordings and heatmaps can provide further clues, subject to appropriate privacy settings. Treat them as observations, not proof of motivation. A repeated click might indicate confusion, a technical failure or simple impatience.

Write down the evidence before proposing the fix. “Several mobile users could not see the validation message” is a stronger starting point than “the form needs a modern design.”

Remove uncertainty before removing every field

Visitors often need more clarity, not fewer words. They may want to know who the service suits, what happens after an enquiry, when delivery arrives or what a quoted price includes.

Answer those questions near the decision they affect. Delivery information hidden several pages away will not help someone hesitating at the cart. A testimonial at the bottom of a long page may arrive too late to support a claim near the top.

Baymard's checkout research documents how checkout design and flow can cause people to abandon purchases. That evidence supports investigating practical obstacles; it does not imply that every abandoned cart can be recovered through interface changes.

Make proof specific to the claim

• A delivery promise needs a clear dispatch and delivery explanation.

• A technical service claim benefits from a relevant work example.

• A quality claim needs product detail or credible evidence.

• A consultation offer should explain what the conversation covers.

Generic praise can sit beside an offer without resolving the visitor's actual doubt. Match the evidence to the decision.

Protect lead quality when reducing friction

A shorter form can generate more submissions. That is helpful only if the added volume is useful to the business.

Consider this hypothetical example, using the same eligible-session count and equal lead-maturity periods:

MeasureOriginal formShorter form Eligible sessions 2,000 2,000 Submitted enquiries 80120 Submission rate 4% 6% Qualified enquiries 4030 Qualified enquiry rate 2% 1.5%

The shorter form appears to win on submission rate. It loses on qualified enquiries and creates more records for sales to review.

That does not prove long forms are better. It shows why the measurement needs to follow the outcome far enough. Test which information genuinely helps qualification and which fields merely create unnecessary effort.

A useful field can save both parties time. An unnecessary field can cost a suitable lead. Decide from the work the field performs.

Build test ideas around a clear explanation

A good hypothesis connects evidence, change and expected behaviour. For example: “Visitors ask about setup time before booking. Showing the expected steps beside the booking form may reduce uncertainty and increase qualified appointments.”

That statement is testable. “Make the page more engaging” is not.

Prioritise ideas using three questions:

1. How many relevant visitors encounter the problem?

2. How strong is the evidence that it affects the decision?

3. What will it take to implement and assess the change?

Keep a short list of high-quality tests. A large backlog of speculative button and colour changes can consume attention without addressing the main obstacle.

Distinguish repairs from experiments. A form that fails to submit valid data needs a fix. A choice between two credible offer structures may need a controlled test.

Run experiments that can support a decision

An A/B test compares variants under a planned allocation. Its value depends on the setup: a defined outcome, sufficient data, reliable tracking and groups that support a fair comparison.

Microsoft Research's work on trustworthy experimentation shows why experiment design and data quality deserve attention before results are interpreted. A dashboard can look precise even when the underlying setup introduces bias.

Before launch, define the primary metric, acceptable risk, minimum meaningful change and analysis approach. Check that both variants work. Keep other major changes out of the test where practical.

Do not repeatedly stop a conventional fixed-horizon test the moment a preferred result appears. Use the stopping rules appropriate to the statistical method you selected.

A small numerical lead is not automatically a reliable winner. An inconclusive test can still tell you that the available evidence does not justify a rollout.

Use a different approach when traffic is limited

Some service websites cannot collect enough qualified conversions for frequent A/B tests. Dividing a small audience across several variants can extend the decision period beyond what the business can use.

In that situation, prioritise clear faults, customer interviews, usability work and substantial improvements supported by evidence. Document the changes and observe later performance with appropriate caution.

A before-and-after comparison is informative, but it cannot isolate the page change from shifts in traffic, seasonality or sales activity. Describe it as an observed change rather than proof of causation.

Keep the scope manageable. Fixing a confusing service description and broken mobile field may be more useful than launching a testing programme the website cannot support.

The discipline is the same: evidence, a clear decision, a documented change and a sensible review.

Build a review that connects marketing and sales

Set a review rhythm based on traffic and the time customers need to make a decision. Give leads enough time to qualify before judging one period against another.

Review the full chain: eligible visits, primary conversions, qualified outcomes and commercial value. Note changes in traffic source and offer mix.

For ecommerce, watch refunds, cancellations and margin alongside orders. For services, review appointment attendance and sales acceptance alongside form submissions.

Record what each test or repair taught you. A losing variant can reveal that a message is unclear or an audience values something different from what the team expected. Keep that learning accessible to the people writing ads and answering sales enquiries.

Frequently asked questions

What is conversion rate optimisation?

Conversion rate optimisation is the process of improving how effectively a website helps suitable visitors complete a useful action. It combines measurement, customer research, practical repairs and testing, with attention to the quality and value of the resulting conversions.

What is a good website conversion rate?

There is no single useful target for every website. Rates vary by offer, traffic source, device, price and conversion definition. Compare like-for-like performance within your business and judge the resulting customers, not just the percentage.

Should I run A/B tests on a low-traffic website?

Only when the available volume can support a useful decision within a reasonable period. With limited traffic, direct usability work, customer feedback and repairs to clear problems may be more productive. Do not describe an under-sampled result as proof.

Improve the decision, then measure the outcome

Conversion rate optimisation works best when it begins with a real obstacle. Clarify the offer, supply the right proof, remove avoidable friction and measure what happens beyond the initial action. The objective is a website that helps the right people become customers on terms that work for the business.

Talk to The Coral Room about conversion optimisation, landing page design and the tracking needed to make better website decisions.