Conversion definition: what counts as a conversion on your site

“Conversion” is a word everyone uses and few teams define twice in the same way. As long as the list of actions counted is not written down, the figures do not match from one dashboard to the next, and the discussion moves onto the tool instead of bearing on the rule.

This page is for the marketing, e-commerce or product director who has to set the definition of a conversion on their site: what counts, in B2C as in B2B, what remains a micro-conversion, and how to trigger it without counting it twice. You will come away with a defensible tagging plan.

What a conversion is called, and why it has to be defined before it is counted

In digital marketing, a conversion reads in operational terms: a visitor moves from an “audience” state to an expected action, measurable and aligned with an objective. Each of these three qualifiers excludes something:

• expected rules out what happens by accident;
• measurable rules out what has to be guessed;
• aligned rules out what serves no stated objective.

That action depends on the context, the channel and the objective set: there is no such thing as a universal conversion, yours is defined before it is calculated. The principle used here, as in our guide to a good conversion rate: the conversion is the action linked to revenue, an order or a qualified contact request; everything else is tracked separately, as a micro-conversion, outside the rate.

That leaves the move from the principle to the list, which is the real exercise. An action deserves to be declared a conversion, the one that counts in the numerator and gets acted on, when it meets three conditions, and all three together:

• it is observable without interpretation: the data alone is enough to say it took place, a confirmed form submission, not an “expression of interest” nor a reading time judged to be long;
• it commits the user beyond a click: it costs them a piece of information, a slot in the diary, a decision. A click on a button commits nothing as long as nothing is sent;
• someone acts when it happens, and that handling leads to revenue: a salesperson calls back, an order ships. An action nobody handles remains an audience indicator.

The third criterion shortens lists: ask, for each candidate action, who receives it and what they do with it. Actions with no recipient drop out on their own, and what remains can be defended in a steering committee.

This page settles the numerator: the event that is counted. The ratio itself, conversions divided by sessions, and the reason why users, clicks or leads would give other values belong to how a conversion rate is calculated. You name the action first, then relate it to the base.

What counts as a conversion in B2C and in e-commerce

On a site, the conversion can be immediate (e-commerce) or gradual (a long B2B cycle). In B2C and in e-commerce, the target action is rarely ambiguous, and the list falls into three levels:

• Conversion: the validated order, the only action that enters the rate;
• Purchase journey micro-conversions: add to basket, payment started, account creation;
• Content and audience micro-conversions: newsletter sign-up, white paper download, webinar registration.

In these models, the difficulty comes less from the list than from the offline part: a conversion can take place by phone or in a shop while being triggered online. Three cases stand apart:

The third case is not a makeshift, provided it is written down that the event tracked is the signal, not the sale it heralds.

What counts as a conversion in B2B, on a long cycle

In B2B, a conversion is not necessarily an immediate sale. It is the qualified contact request, the stage of the sales cycle that implies a project: a demo request, a quote request or a meeting booked. The other useful actions in the journey are tracked as micro-conversions, outside the rate:

• white paper download;
• webinar registration;
• account creation;
• generic contact form sent.

The reason is arithmetical as much as commercial. Weeks or months pass between the first visit and the signature, and the decision is not taken on the site. Waiting for the sale would produce a series of a few units a month, in which no variation means anything. So the most advanced stage that remains observable online is used.

Two practical consequences follow. First, the generic contact form and the demo request are not worth the same: the second implies a project, the first may come from a job applicant or a supplier. Separating them at the tagging stage avoids sorting by hand in the sales tool. Second, the qualification decided off the site (lead accepted, meeting attended) lives in the CRM, under a rule of its own, and not in the tagging plan.

Macro and micro-conversions: two levels, one steering

In practice, several actions are tracked, ranked by value: the macro-conversion reflects the business, micro-conversions reflect progress through the journey. The top level on its own leaves steering blind between two signatures; the bottom level on its own gives the illusion of movement where nothing is being sold.

The rule that links the two fits in one sentence: a micro-conversion only means something if it is linked to a higher probability of macro-conversion. If a visit to the pricing page often precedes a demo request, that visit deserves to be tracked as a milestone. If, conversely, an action occurs just as readily among those who buy as among those who leave, it is not a micro-conversion: it is an audience measurement.

The macro-conversion: the action that carries the business value

A macro-conversion corresponds to the priority final objective, linked to revenue: an order, a quote request, a demo request, a meeting booked. It, and it alone, enters the conversion rate; it serves to measure direct business performance and to calculate a return on investment.

Three rules frame the choice:

• one per business model: as soon as three are declared, nobody knows what is meant any more by “conversions have gone up”;
• distinct, named actions: an online order and a quote request on the same site are never added into a single total whose rise cannot be read;
• a deciding criterion: the action beyond which the company considers a commitment to have been made.

Micro-conversions: the intent signals

Micro-conversions do not meet the three conditions set out above, and that is not a flaw: they are measurement milestones, not actions to be handled. Each one signals growing intent, none triggers a sales call-back or enters the published rate:

• a visit to a pricing page;
• a click on a CTA;
• a form start;
• an add to basket;
• a document download;
• a newsletter sign-up;
• a visit to a key section.

These milestones serve the diagnosis. They are particularly useful for:

• spotting friction ahead of the final objective;
• measuring progress when the macro-conversion barely moves;
• segmenting by intent (top of funnel vs bottom of funnel).

Their number calls for discipline. Five to eight micro-conversions cover an ordinary journey; beyond that, the plan becomes impossible to maintain and the events overlap: a “CTA click” and a “form start” often count the same gesture. For each one, ask who looks at it and which decision it serves.

A tagging plan is summed up in a table with four entries: the action, its level, the stage of the journey, and what triggers the event. It is that document, and not the configuration screen, that carries authority on the day a figure is contested.

Where each conversion sits in the funnel

A funnel describes the stages from arrival on the site to the final action. A simple reading in four levels:

• Discovery: the user arrives via a query, a piece of content or a campaign.
• Consideration: they compare, look for evidence, explore.
• Conversion: they complete the expected action.
• Activation: they get the value (onboarding, first use, first commercial response, and so on).

Place the intent signals as progress milestones and the high-value action as the final objective: what you get is a measurable funnel, useful for diagnosing where things get stuck. Every declared action attaches to one stage and one only; placed in two spots, it produces two contradictory readings of the same journey.

The right reflex is then to measure the losses between stages (e.g. pricing page visits → form start → submission). A clear drop at one stage points to a place, not a trend, and is easier to handle than an overall rate. The two stages compared must still cover the same population and the same period, failing which the drop is only a mismatch of scope.

When the loss sits on the payment funnel or on a form that misbehaves, the subject becomes an optimisation project, shared between product, UX and development as the good conversion rate guide explains. The tagging plan has done its job by showing the exact spot.

Tagging a conversion in the measurement tool

A declared action only becomes a figure if the tool recognises it, once and once only. This is where plans come undone, in three ways:

• the event fires twice, on a reload or a re-submission;
• it fires too early, on a click rather than on a confirmed submission;
• it does not fire at all for part of the visitors.

This plan says which actions fire, and under which conditions. Breaking the result down afterwards by channel, entry page, device or campaign belongs to a separate setup, described with the conversion rate KPIs.

Defining the event: a single action, observable, documented

In GA4, a conversion corresponds to an event that you mark as a key event. The quality of the measurement therefore depends on your tagging plan:

• define a single, observable action (e.g. a confirmed form submission);
• avoid ambiguous events (e.g. a click on a button with no evidence of submission);
• document the rule (when is the event triggered? on which page? with which exclusions?).

The configuration follows the same requirements, to be checked one by one before publishing any series:

• a stable event, consistently named and triggered once;
• a server signal rather than the mere display of a confirmation page, which repeats on every reload;
• a transaction or submission ID sent with the event, the only way to deduplicate a purchase;
• separate handling of cancellations and refunds, which remove a conversion already counted;
• tests before going live, on mobile and on desktop.

On naming, two events naming the same action under two labels are the most ordinary cause of a gap between two dashboards. Reading these events in the reports, once the plan is in place, is covered in the conversion rate in Google Analytics.

Checking the data: consent, cross-domain, deduplication

Before interpreting a series, check three risk areas:

• Consent: refusals can under-measure some conversions, depending on your implementation.
• Cross-domain: payment, meeting booking or hosted forms can break the session if badly configured.
• Deduplication: reloads, back navigation and double submissions are deduplicated on a technical submission or transaction ID, never on an email address, which is personal data and has no business being sent to an audience measurement tool.

Three reading errors complete the list, and those are corrected by no setting:

• Duplicates: the same lead counted twice (a return to the confirmation page, a re-submission).
• Sample bias: concluding that there is a rise after an atypical week or a one-off campaign.
• Ignored break in series: comparing two periods on either side of a change of tagging rule.

These checks are recorded with the list of actions, and dated: without a date, it is impossible to know whether a break in a series comes from the market or from a change of rule. Any change of definition or tagging opens a new series, kept alongside the old one without recalculating it.

What counts as a conversion when part of the journey happens before the click

A growing share of the journey takes place “before the click”: the search ends without the user opening a single result, the answer being delivered in the interface. One data point gives the order of magnitude, and nothing more: the annual growth in global traffic referred by generative AI platforms is estimated at +300% (Coalition Technologies, 2025), according to our GEO statistics.

For a tagging plan, the consequence is precise: part of the conversions will arrive with no prior journey, because the comparison took place elsewhere and the visitor lands directly on the page that allows them to act. The consideration signals then disappear, not because interest is falling, but because the stage played out beyond the tool's reach. Reading that disappearance as a drop in intent is the mistake of the moment.

Two decisions follow, and each one is written into the plan. The table sorts actions by what they still allow you to measure:

The first decision separates the first two rows: actions that assumed a complete journey keep their diagnostic value, but stop serving as the reference measurement. The second is a limit to be written down: without an origin signal (a campaign parameter, a referrer passed on, an identifier kept end to end), a direct arrival remains direct traffic. Assuming it came from a generative answer is a hypothesis, and it is not published as a figure.

The consequence for content: the journey shortens, and the requirement for clarity grows accordingly. A page that serves as a source for a generative answer without offering any observable action will produce no measurable conversion: keep an explicit call to action and value proposition on it.

Continue reading

Depending on where you stand once the tagging plan is written, the next step is in one of these articles:

• Your funnel shows where it breaks down: if the loss is confirmed over several periods, the subject becomes a prioritisation of pages and a measurement of effect, covered in optimizing conversion rate.
• You are looking for which conversions to attach to the organic channel: if part of the journey goes through generative answers, attribution becomes a project in its own right, described in what SEO does to conversion, and how to attribute it.
• You are judging a page whose only function is to convert: if its result differs clearly from the rest of the site, it is the page that has to be read, as explained in the conversion rate of a landing page.

Once the tagging plan is written, tracking the declared micro and macro-conversions together, and seeing at which stage the funnel loses people, means bringing them into a single view: that is what the performance reporting module centralises.

FAQ: defining and tagging a conversion

What is the definition of a conversion in digital marketing?

In digital marketing, a conversion is the expected action a user completes, linked to revenue and measurable through an event in an analytics tool: an order, a demo request, a quote request or a meeting booked. It must be observable without interpretation, commit the user beyond a click and trigger handling.

What are the different types of conversion?

There are two levels: the conversion, linked to revenue, and micro-conversions. In e-commerce, the conversion is the order; in lead generation, it is the demo, quote or meeting request. Add to basket, account creation, sign-up, download or visits to strategic pages are micro-conversions, outside the rate.

What is a “conversion” in B2B?

In B2B, the conversion is the qualified contact request: a demo request, a quote request or a meeting booked. The most advanced stage that remains observable online is used, since the signature plays out off the site and over several months. White paper, webinar, account creation and generic contact form remain micro-conversions.

What is the difference between a micro-conversion and a macro-conversion?

The macro-conversion is the final objective linked to revenue (order, demo, quote) and the only action that enters the rate. The micro-conversion is an intent milestone (pricing page, form start, add to basket) that helps steer the funnel without any sales handling. It is only justified if it signals a more likely macro-conversion.

How is a conversion rate calculated?

The conversion rate is calculated as follows: (conversions / sessions) × 100. The numerator only counts the action linked to revenue, never micro-conversions; the denominator stays sessions, because users, clicks or leads would give another value for the same month. Any change of definition opens a new series.

How are conversions tracked with Google Analytics 4 (GA4)?

In GA4, create an event (e.g. a validated form_submit), mark it as a key event and check that it fires in debug mode, on mobile and on desktop. Then check consent, cross-domain and deduplication before interpreting any series at all: a measurement gap can be corrected, a badly tagged series has to be thrown away.