Conversion rate KPIs: the indicators to track and the breakdowns that explain the gap

This page is for the marketing, e-commerce or product director who has to present a conversion rate to a committee and make it steerable. The conversion rate KPI decides nothing on its own: what turns it into a decision tool is the indicators read alongside it, the breakdowns that explain its movements and the reliability of the measurement.

The criteria that make your rate comparable to a benchmark, from the denominator to the page type, are set out in what makes a good conversion rate. This page starts from there: what to track, how to break it down, and how often to review it.

A conversion rate is only a KPI if it is tied to an objective

A KPI (Key Performance Indicator) measures how effective a strategy is against an objective decided upstream. It only makes sense if that objective is relevant: precise, realistic, achievable and time-bound. If the objective is revenue growth, the matching KPI can be sales volume; if it is the visibility of the site, organic traffic.

In digital marketing, the conversion rate is central, because it shows the ability to turn visitors into customers or prospects.

The rule works the other way round too, and that is where it earns its keep: an indicator that no decision is waiting on has no place on a dashboard. Before adding a line, ask yourself these questions:

• Which decision does it support?
• Who takes that decision?
• By when does the decision fall?

Indicators that answer none of the three belong in an analysis report. That sorting is what separates a leadership dashboard from an export out of a measurement tool.

You still have to agree on what the rate measures: which action counts as a conversion, and which population it is divided by. The definition of a conversion rate and of its denominator settles this once and for all: conversions divided by sessions, with micro-conversions tracked separately. Until it is settled, two departments in the same company present two different rates for the same month, and the meeting turns on method rather than on performance.

Once that rate is set, four elements make it usable:

• Neighbouring indicators: those read alongside it, which give it a sign.
• Breakdowns: those that explain its movements.
• Generative indicators: those that track the brand's presence in AI answers.
• Measurement checks: those that stop you commenting on a measurement artefact.

The indicators read alongside the conversion rate

A rate on its own says that something happened, never what it is worth. A rate that rises while cost per acquisition rises faster is bad news; a stable rate on a volume that has halved should not be commented on at all. It is the neighbouring indicators that give the sign.

To connect the rate to a value, a floor of three lines is already enough to settle an acquisition budget without settling it blind:

• Conversions by channel: SEO, SEA, direct, email.
• CPA or value per session: what a conversion costs and what it brings in.
• Pipeline contribution: the MQLs and SQLs produced, in B2B.

Beyond that floor, each indicator decides something different, and it is that mapping that has to be fixed before building the dashboard. Three lines call for a written definition; without one, they get misread:

• Conversion rate: conversions divided by sessions, never by users.
• Revenue per session: revenue divided by sessions, the same denominator as the rate; calculated per user, it would give a different value, and the gap would widen with repeat visits.
• Bounce rate: in GA4, the share of non-engaged sessions; a single-page session lasting more than ten seconds is engaged, so it does not bounce.

The table matches each indicator to the decision it carries:

Measuring funnel abandonment is part of running measurement; fixing the funnel belongs to the product and development teams, on their calendar. The indicator documents the request and makes it arbitrable; it does not handle it.

One rule applies to the list itself. The reflex is to lengthen it, every question raised in committee producing one more line, and the dashboard becomes unreadable within two quarters. Settle on a short list, date its last change, and extend it only against an expected decision.

Segmenting the rate: where to cut it, and what each breakdown reveals

An overall rate is an average, and an average is almost always the wrong level to read at. Segmenting avoids misleading averages and points to the quick wins, for example a heavily visited SEO landing page that converts little. Each breakdown has its function and its place in the reading:

The order matters: the channel opens the reading because it is the only breakdown whose composition changes without anyone having touched the site. Device, audience and campaign refine a diagnosis; they rarely establish one.

By acquisition channel and by campaign

A gap by channel says something about the mix first, rarely about the site: if organic and SEA convert at a stable level and the overall rate falls, then the share of each has changed. So compare each channel's rate to itself across the two periods before commenting on the total.

A campaign is read separately from its channel, because it carries its own promise and its own target. What each channel brings in and what it costs across a full mix is covered by the digital marketing conversion rate by channel.

By entry page, by device and by conversion type

The entry page is the test of the promise. A heavily visited page that converts little nearly always signals a gap between what the title announces and what the page delivers, or between the intent behind the query and the offer presented. It is the breakdown that produces the most immediate decisions: it names a precise object to fix, not a trend to watch.

Device is a signal, not a verdict. A mobile/desktop gap on the same page reads first as a difference in usage context, and only then as an interface fault. When it really is the interface, the fix belongs to the product, UX and development teams; the indicator quantifies the request.

Finally, conversion type separates the actions tied to revenue, the order and the qualified contact request, to see the weight of each. Micro-conversions such as a sign-up or a download are tracked separately: they describe the journey, they do not enter the rate.

Conversion indicators coming out of generative engines

GEO adds three lines to the dashboard, to be tracked separately from organic: merged into organic, a population that behaves differently pulls the average without anyone knowing in which direction.

• Generative share of voice: how often your pages are cited in AI answers, measured against a list of phrasings settled in advance and read in its trajectory from one reading to the next.
• Sessions coming from generative answers: tracked by volume, because the impressions of a citation are reported nowhere and the denominator of a click-through rate does not exist.
• GEO attribution: the link between cited content, sessions and conversions; until it is built, the first two lines remain visibility indicators, not a contribution to revenue.

These indicators are young and their definitions vary from one tool to another: track them as trends on your own scope, and read the values published in the GEO statistics as outside observations that you will not reproduce.

Making the measurement reliable before interpreting it

The most frequent case in committee is not a rate that degrades, but a rate that moves when nothing has moved. Before looking for a business explanation for a movement, check that the movement exists: a gap smaller than what the measurement can resolve is not worth commenting on.

That check assumes you know how the rate is produced: which events are tagged, which dimensions cross, which filters apply. Setting it up and reading it inside the tool is covered by the conversion rate in Google Analytics.

The five errors that move a rate when nothing has moved

When a rate drops off, each error shows up at a specific point in the measurement:

The last is the most expensive, because it produces plausible conclusions: a holiday month, an SEA budget cut, a redesign shipped mid-month explain a movement long before page quality does. This drop-off check comes before any comment, after every release as at every review.

What gets documented, and what gets revalidated after every release

Making it reliable means documenting events, validating tracking after every update and aligning marketing and sales on a shared nomenclature. These actions only hold over time in the form of lasting deliverables:

• An event register: name, trigger, page concerned, creation date, owner.
• A non-regression sign-off: after every release, the main conversions are replayed and checked to report once and only once.
• A shared nomenclature: the same words for the same objects across marketing and sales (visitor, lead, MQL, SQL, customer), so that a rate presented by one is read without translation by the other.

The owner matters as much as the document. Without a name against each line of the register, revalidation happens after the incident. A quarterly review works, provided someone is named responsible.

From the observed gap to the decision

Once the measurement is reliable and the rate is broken down, place the gap against three references, always in this order:

• Your history: twelve months at least on a constant definition, read on a rolling monthly basis; it is the only reference whose calculation method you know end to end.
• Your segments: the gap between two of your channels or page types points to potential more reliably than a gap against a published average.
• Published benchmarks, last: on an identical denominator, as an order of magnitude, never as a target.

Before using a benchmark, check what the source measures: which action, which denominator, which population. Competitors are not on the list, as their rates are not accessible. Also adapt the analysis to your business by reading the rate with other KPIs, such as average order value or session volume: a rate that rises on a volume that falls does not produce more revenue.

The review cadence

Review frequency is set by the nature of the decision, never daily:

Looking at a rate every day means commenting on noise. The quarter is also the first period over which a content change has had time to produce a measurable effect. The cadence says when to review; the decision window, for its part, is set by the volume of conversions observed, not by the calendar, as conversion rate optimisation explains in detail.

What a gap allows you to conclude

A confirmed gap allows you to say three things, already missing from most dashboards:

• Where the rate drops off: the channel, the page, the device.
• By how much: the size of the gap, relative to the volume concerned.
• Since when: the break date, to be checked against the release calendar.

It does not allow you to name a cause: two indicators moving together establish no link between them, and a single period does not make a trend. Moving from observation to action, knowing which page to start with, what to change on it and how to measure the effect, is a matter of prioritisation, which assumes a diagnosis and not a reading.

One last reflex: the day you settle your dashboard, write down the exact definition of each line, denominator included. Any change of definition or tracking then opens a new series, kept alongside the old one. That is what will stop you, six months later, comparing two figures that no longer measure the same thing.

Continue reading

Depending on what reading your KPIs has brought to light, the next step is not the same:

• Your organic traffic mixes SEO and generative answers: building the attribution model then changes in nature, as explained in what SEO does to conversion, and how to attribute it.
• A segment drops off over several periods: the subject becomes page prioritisation and effect measurement, covered in optimizing the conversion rate.
• Your business is seasonal or highly specialised: an order of magnitude can be read, on an identical denominator, in the conversion rate by business sector.

To read the rate segment by segment without rebuilding the export by hand, the performance reporting module centralises SEO and GEO KPIs for unified steering of your conversion rate.

Frequently asked questions

Which conversion rate KPIs are essential?

The essential conversion rate KPIs are the rate itself, overall and by channel, cost per acquisition, revenue per session, bounce rate by entry page, average session duration, funnel abandonment rate and customer lifetime value. Generative share of voice is added to track GEO. Each line must answer an expected decision.

How often should conversion KPIs be analysed?

Conversion KPIs are reviewed weekly for alerts, monthly for decisions and quarterly for trends and budget trade-offs, never daily. The weekly review spots an anomaly without concluding. Automated alerts save you waiting for the monthly review to see a break in collection.

What is the difference between a conversion rate KPI and a conversion?

A conversion is the action completed, such as a purchase or a demo request; the conversion rate KPI measures the share of sessions that complete that action. One is an event, the other an indicator built on that event. What matters is defining clearly the action counted, the scope and the period.

What is a good conversion rate KPI in B2B?

A good conversion rate KPI in B2B is a rate that improves against your own history, on the same segment and a constant definition, while staying profitable given the cost of acquisition. Published benchmarks come last, on an identical denominator, as an order of magnitude. Intermediate conversions, such as a download, are tracked separately from the rate.

How do you segment the conversion rate KPI to analyse it properly?

The conversion rate KPI is segmented by channel, entry page, device, audience and conversion type. Open with the channel when you do not yet know where to look, because it is the breakdown whose composition changes without any intervention on the site, then drill down into the entry page, which names the object to fix.

Which common errors distort conversion rate KPI tracking?

Five errors most often distort tracking: a badly tagged event, double counting, mixing users and sessions, internal traffic not excluded and reading without context. An event register, a check after every release and a shared nomenclature across marketing and sales prevent them.

How do you connect conversion KPIs to SEO/GEO ROI?

Connecting conversion and ROI means linking each conversion to its source, SEO or GEO, then to a value: revenue, margin or lead value. Track at minimum conversions by channel, CPA or value per session and pipeline contribution in B2B. On the GEO side, add generative share of voice and the attribution of cited content to conversions.