Optimizing the conversion rate: which page to start with, and how to read the effect

You have three hundred pages, a quarter ahead of you and a trade-off to settle between improving what exists and producing something new. Conversion rate optimisation then raises three questions: which page to start with, on what criterion to defend that order, and how you will know in two months that the change produced an effect. This is a matter of method and steering, not a list of recipes.

It assumes you know what your rate compares with: denominator, channel, sales cycle, offer and page type are set out in what makes a good conversion rate.

What optimizing a conversion rate means, and what it involves

A conversion is a visitor to a website doing what is expected of them, in the form of an action linked to revenue. Depending on the type of site or the line of business, it takes one of these forms:

• A sale, most of the time, in e-commerce.
• Starting a free trial, if you sell a subscription.
• A qualified contact request, when the sale is closed off the site.

Signing up to a newsletter or watching a video are micro-conversions: they are tracked separately and do not count in the rate, which is calculated as conversions divided by sessions.

What a point of rate is worth shows on an example, for a site with 200,000 monthly sessions:

• Before: a conversion rate of 3%, that is 6,000 conversions.
• After: a rate of 4%, that is 8,000 conversions.
• The gap: two thousand more conversions, without a single extra session.

It is that ratio which justifies devoting a quarter to it.

The word "optimize" is misleading, though. It suggests it would be enough to run through a list of things to do on a page. The real problem is a problem of order, which raises three questions:

• which pages to work on first;
• with how much effort;
• how to know the effort paid off.

The techniques are known; what is missing is the ranking.

The method is therefore about order and control, on a scope better named straight away: the editorial optimization of organic entry pages. It runs in five stages:

• the inventory of candidate pages;
• the criterion that ranks them;
• the page-by-page decision;
• the examination that comes before any change;
• reading the effect.

Defects in forms, in the checkout funnel and in loading time stay in the overall backlog, quantified and dated, even when carrying them out falls to other teams.

Building the inventory: which pages can move

Before any ranking there has to be a list, which is not drawn up on intuition. It is produced by crossing three readings you already have:

• the pages that receive organic traffic;
• those whose entry intent is close to a conversion;
• those whose rate is below that of their peers.

A page that fails one of the three leaves the inventory, not for good, but for this quarter.

The most profitable entry rule is simple. Start with the pages that combine good rankings and a low conversion rate: these are the pages with the greatest untapped potential, since they are already visible but do not convert enough. The gap plays out between what the page promises and what it delivers.

Two families of pages turn up at the inventory without belonging there; the table separates them from the pages the method handles.

The first may have an excellent rate: it simply receives nobody, and optimising click-through rate in the SERP handles that upstream link.

The second is scheduled with the product, UX and development teams: note the page, quantify the shortfall and leave the line open in the overall backlog with an owner and a date. The full division of responsibilities is set out in conversion rate optimisation and the division of roles it assumes.

Prioritizing: the scoring that produces the backlog

Once the inventory is made, it has to be ordered, and the order has to be defensible in front of someone who did not do the survey. Prioritize the pages that combine five criteria:

• High organic traffic, stable over the period.
• Intent close to conversion: landing pages, product pages, pricing.
• A lower rate than that of their peers.
• Direct business impact: qualified leads, MRR, average order value.
• An easy fix, one that fits within the content.

A simple scoring condenses them to build the backlog: traffic × value × ease, with value bringing together intent, the rate gap and business impact. Traffic and value rank pages by stake; ease ranks them by cost. Without it, the backlog pushes up the heaviest pages, the ones that will never ship, and the quarter ends with nothing delivered.

Each of the five criteria is recorded in a precise place: that is what makes the order discussable rather than arbitrary.

The intent criterion does not point to the same pages depending on the market. In B2B, the conversion is often a demo request or a meeting booked rather than an immediate purchase; a download, for its part, is tracked separately as a micro-conversion. The criterion there brings up comparison pages, evaluation guides and offer pages, where a retailer would bring up product pages. Set what counts as your conversion before ordering.

Deciding whether the page is worth the effort: the value of a visit

An ordered backlog does not yet say whether it should be worked on. A page can rank first and deserve nothing, because what a visit is worth on it does not cover what reworking it would cost. The question is settled with a single quantity.

What a visit brings in

Suppose that in one month your site receives a hundred sessions, for a conversion rate of 5% and an average order value of €50. The calculation runs in three steps:

• The value of a visit: (5 × 50) / 100, that is €2.50 per session.
• The incremental margin gain: the extra conversions you are aiming for, multiplied by the margin each one yields, not by the revenue.
• The comparison with cost: that gain, and that alone, is compared with the cost of the fix under consideration.

On this example, aiming for one more point of rate comes down to one more conversion a month. If the margin per sale is €15, the fix is only defensible if it costs less than what it will bring in over the period it stays in place.

It is the same quantity as the value per visit (RPV, revenue per session) tracked on the dashboard, and it is calculated page by page, never for the whole site: a product page and an in-depth article have neither the same conversion value nor the same acquisition cost. Note it in the backlog next to the score, recorded over the same window as the rate.

Optimizing what exists or going after traffic

Each combination of visit value and volume calls for a different decision:

The page with high value and high volume, contrary to intuition, is not the one you open first: it carries a visible share of the result, and a change there risks more than it promises.

For the others, the rule is the incremental gain set against the cost: if the expected additional margin exceeds the cost of the fix, and exceeds it more surely than the same budget placed in acquisition, you rework the page. This calculation is redone every quarter, because both terms move.

What to look at before touching the page

The page is chosen, the effort approved. One examination comes before all the others, the one that explains most content gaps. It covers three things that have to agree:

• what the entry query asks for;
• what the page announces in its title and first lines;
• what the offer actually delivers.

When these three move apart, no work on form makes up the mismatch.

Misalignment can be read without tools. Record the queries that bring people to the page, read them as a question, then read the page as an answer to that question. If the question is "how do I choose" and the page sells, the visitor leaves. If the question is "how much does it cost" and the page tells a story about an approach, they leave too.

The fix is either to realign the page on the intent it receives, or to move the intent to another page of the site, depending on what the rest of the site already covers.

This work assumes you know where the traffic comes from and what it does. Analytics tools serve to understand visitors' habits and interests:

• which sources generate a conversion most easily;
• how the different target segments interact with the content;
• on which devices they do so.

The tool does not convert, it points to where to look: without that reading, the alignment check is done at random.

Finally, a page ranking well on a precise query receives visitors who know what they are looking for, and its expected rate is not that of a page fed by a broad query. Position therefore counts in the order of the backlog.

Measuring the effect: the before, the window, non-regression

An optimization whose effect is not read is not finished, it is only delivered. The setup rests on three actions, taken before changing anything at all:

• The initial reading, dated and filed in the backlog.
• The comparison window, set on the volume of conversions.
• Non-regression, checked when the effect is read.

The initial reading

Before touching the page, note, over a reference window long enough to absorb week-to-week variation:

• its conversion rate by source;
• its average position;
• its volume of entries;
• its value per visit.

Place that reading within the page's history over at least twelve months, to locate seasonality. Dated, it alone will serve as the point of comparison: a before reconstructed after the fact is not a before. If the definition of the conversion or the tagging changes along the way, a new series opens, kept alongside the old one without recalculating it.

The comparison window

The decision window is set by the volume of conversions, not by the calendar. Compare periods of equal length, each containing enough conversions for a gap to be readable, and set that threshold before changing anything at all.

A properly filled-in backlog line, for a pricing page, looks like this:

• Hypothesis: the page mostly receives the query "how much does it cost" and displays no order of magnitude before the form, which would send part of the visitors away.
• Volume: 4,000 organic entries a month, 80 quote requests, that is 2%.
• Minimum effect sought: half a point, that is twenty more requests a month; below that, the gap will not be distinguishable from this page's ordinary variation.
• Cost: two days of copywriting, half a day of integration.
• Risk: displaying a price range may put off small accounts, so the qualified share of the requests is recorded alongside their number, and a rollback stays planned.
• Protocol: a dated initial reading covering the set conversion threshold, publication, then a period of the same length at a constant mix of sources, with no other change on the page.
• Expected result: 100 requests a month instead of 80, with a qualified share at least equal.

Written this way, the line is reread in two months with no discussion about what was being looked for.

Non-regression

When reading the effect, check that nothing else has moved between the two periods:

• same mix of sources;
• same seasonality;
• no concurrent redesign;
• no campaign launched or stopped in the meantime: a point gained while a paid campaign is being switched off goes to nobody's credit.

The reading, finally, assumes the rate is already broken down. How to segment that rate, by channel, entry page, device or campaign, and how often to read it are handled with the conversion rate KPIs; here, segmentation is a prerequisite, not a subject of work.

Generative engines change nothing in the method, but they alter the ranking: a page already cited in their answers often receives traffic with stronger intent, to be checked on the page, and moves up at equal score. A page that answers precisely the question asked serves both conversion and citation.

Continue reading

Depending on what is blocking the reading of the effect in your case, the next step is not the same:

• You do not know what share of your conversions belongs to organic: until attribution is settled, the effect of an optimization stays unreadable, as shown in what SEO does to conversion, and how to attribute it.
• Your backlog mostly brings up campaign pages: designed to convert, they do not expect the same levels as a product page or an article, as detailed in the conversion rate of a landing page.
• Your two dashboards do not show the same rate for the same page: no before-and-after will hold until the denominator is settled, which is covered by the definition of a conversion rate and what is divided by what.
• The tagging is not in place: you need to instrument before comparing a before and an after, starting with what the Google Analytics measurement tool lets you track.

To track a page's rate before and after optimization, segmented by source, the performance reporting module centralizes conversion, SEO and GEO metrics in a single dashboard.

Frequently asked questions

How do you optimize a site's conversion rate?

You optimize a conversion rate in three stages, always in this order. Build the inventory of pages that receive traffic, carry an intent close to conversion and convert less than their peers. Order them on a traffic × value × ease scoring. Record the initial state, make the change, then compare over an equal window.

Which actions have the most impact on the conversion rate?

The action with the most impact is alignment between the entry query, the page's promise and the offer. Next come a clear value proposition, visible CTAs and reassurance through proof and reviews. Loading speed and simpler forms often count just as much: they are handled with the product team and stay in the overall backlog, quantified.

Which tools should you use for conversion rate optimisation?

Two tools are enough to hold the method. A tool that measures the rate by page and by source is used to build the inventory, record the initial state and read the effect. Search Console covers the upstream side: entry queries by page, impressions, clicks and average position.

How long before you see the results of an optimization?

The duration is set in volume of conversions, not in weeks. Decide before making a change which gap you want to distinguish, then wait until each period contains enough to read it: on a page producing twenty conversions a month, half a point takes several months; on a page producing a thousand, a few weeks.

How do you prioritize the pages to optimize?

You prioritize the pages that combine high traffic, intent close to conversion, a rate below that of their peers, direct business impact and an easy fix. A traffic × value × ease scoring puts them in order. Start with those that combine good rankings and a low rate: they are already visible and do not convert enough.

Which common mistakes push the conversion rate down?

On the content side, the common mistakes are a vague message whose value proposition is not visible, CTAs that are not explicit, a lack of reassurance and too many ways out. On the interface side, forms, slow pages and mobile experience belong to the product and development teams. The most expensive remains a decision based on best practice with no measurement.

How does conversion rate optimisation apply in B2B?

In B2B, the conversion is often a demo request or a meeting booked rather than an immediate purchase. Optimization runs through CTAs matched to the level of maturity, progressive forms and strong alignment between SEO content and the offer.