What Is a Conversion? Conversion Rate Formula and GA4 Setup

A conversion is an action a visitor takes on your website or app that your business has decided counts as a result: submitting a quote request form, completing a checkout, tapping a call button, or clicking through to add your LINE Official Account. Conversion rate is the share of opportunities that turned into one of those actions, usually conversions divided by sessions, users, or ad clicks, expressed as a percentage.

There is no universal definition, because every business picks its own, and every platform counts at a different point over a different time window. This guide covers what a conversion is in practice, how the rate is calculated, where to set it up in GA4, why the Google Ads number never matches the GA4 number, and why looking up the answer to "what is a good conversion rate" sends most people in the wrong direction.

What a conversion is in digital marketing

Technically, a conversion is one event you have declared to be commercially meaningful. No measurement system works out on its own that a form submission matters more than a scroll. A person configures that. So a conversion is a business policy decision, not a property of the website.

The consequence is that two companies in the same industry can report numbers ten times apart purely because of definitions. Site A counts only paid orders. Site B counts every tap on a chat button. B looks far better while selling less. When someone quotes their conversion rate, the first question worth asking is what they counted.

A workable definition passes one test: if this number doubled next month, would the business actually take in more money? If the answer is not clearly yes, the thing being counted is a signal along the way rather than a primary conversion.

Macro and micro conversions

A macro conversion maps directly onto a business goal: a completed payment, a quote request, a booked appointment. A micro conversion is a step on the way that shows someone moving closer: reading the pricing page to the end, downloading a brochure, revealing a phone number, starting a form without finishing it.

Both layers exist because of data volume. A B2B company winning ten new clients a month cannot analyse anything from macro conversions alone, since the counts are too small to separate a real change from chance. Add micro conversions and you can see which step people drop at and which page they read before leaving.

The caution is that micro conversions are for diagnosis, not for automated bidding to chase. Set a chat button tap as a campaign primary goal and the system will go looking for people who like tapping chat buttons, who are not always the people who like paying.

The table below compares how the two layers are counted, along with what each is safe to use for.

How conversion rate is calculated, with a worked example

The formula is conversions divided by a base unit, times 100. Everything hinges on the base unit, because each tool picks a different one.

• GA4 reports session key event rate, which divides by sessions, and user key event rate, which divides by users. The two are rarely equal, since one person can open several sessions.
• Google Ads divides conversions by interactions, which for a Search campaign means clicks, not sessions on the site.
• Sales teams usually divide closed customers by leads accepted into the CRM, a third base unit again.

The numbers in the example that follows are illustrative only. They are there to show the arithmetic and are not measurements from any real site. Say a site records 12,000 sessions in a month and 240 form submissions. The session level rate is 240 divided by 12,000, times 100, which is 2 percent.

That single figure says very little until it is split. Suppose 9,000 of those sessions came from mobile and produced 108 forms, a rate of 1.2 percent, while 3,000 desktop sessions produced 132 forms, a rate of 4.4 percent. The blended 2 percent hides the fact that the mobile form has a problem, which is the part you can act on.

One more layer gets forgotten: the paths that do not end on the site at all. If the same month produced 600 clicks out to LINE, counting only forms shows roughly a third of the contact intent that actually occurred. Plenty of Thai sites fall into this, because a form is not the channel most Thai buyers reach for.

What a good conversion rate is, and why published averages mislead

The direct answer is that there is no benchmark you can judge yourself against, and comparing your site to a number from an article usually leads to a bad decision. The problem is not that the published number is false. It is that it measures something other than what you measure.

Several layers stack up here. First, definitions differ, as described above. Second, the base unit differs: a rate calculated on ad clicks will always sit higher than one calculated on all site sessions, because people who click an ad arrive with more intent than average traffic.

Third, traffic mix matters. A site whose visitors mostly search for its own brand name will show a higher rate than one fed by educational articles, even when the pages are equally good. Fourth, price and decision cycle: a six-figure baht purchase that takes three months of discussion cannot produce rates comparable to a 200 baht impulse buy.

What works instead is a conversion rate baseline of your own. Build one like this:

1. Fix the conversion definition first, and leave it alone for the whole collection period.
2. Collect at least 8 to 12 weeks of history, so it spans both payday weeks and holiday periods.
3. Calculate the rate separately by device, channel, landing page, and campaign, rather than keeping one blended figure.
4. Look at weekly medians rather than a period average, because a promotional week drags an average badly.
5. Record every site change, budget change, and tag change, since those explain swings more often than seasonality does.

Once the baseline exists, the question changes from "is this below the industry standard" to "is this week far enough from the site's own normal to be worth investigating", which is answerable with data you actually hold.

Sample size deserves a warning. If a month produces conversions in the single or low double digits, a move from 2 percent to 3 percent can come from a handful of extra events and cannot be read as a trend. In that situation, always read the raw counts alongside the rate, and lean on higher volume micro conversions to judge direction.

How to set up conversions in GA4

GA4 uses the term key event for the important actions measured inside the property, and reserves the word conversion for figures passed to Google Ads. The rename confuses people opening the reports for the first time, since the old Conversions menu is gone even though the mechanism underneath is the same.

To mark an event GA4 already collects:

1. Open Admin, the gear icon at the bottom left.
2. Under Data display, choose Events, which lists the events collected in the past 30 days.
3. Find the event you want and switch on Mark as key event in the right hand column.
4. If the event has never fired yet, go to Key events and create it using exactly the name the site will send.
5. Verify in the Realtime report or DebugView that the event fires once per action rather than twice.

If the event does not exist, build one from the Events menu with Create event, using a condition based on an event you already collect. For example, create generate_lead when a page_view occurs whose page_location contains thank-you. That works only when the form really navigates to a thank you page. If the form submits without a page change, the event has to fire from Google Tag Manager or from the site code.

GA4 enhanced measurement already collects several things automatically, including scrolls, outbound clicks, site search, video engagement, and file downloads. The outbound click event is useful on Thai sites, because it captures clicks out to LINE or to a marketplace storefront without any extra tagging.

Two settings deserve a deliberate decision early. In GA4 attribution settings, the lookback window for other conversion events can be set to 30, 60, or 90 days, with 90 as the default. Separately, data retention for user level data on a standard property can be set to 2 months or 14 months. That setting affects Explorations, which query the underlying data, and not the prebuilt aggregated reports. Teams that leave the default and later ask for last year's detail usually find it has already expired.

For a full measurement rebuild, from event structure through to the Google Ads connection, see the Google Analytics 4 setup and migration service.

Why Google Ads and GA4 conversion numbers never match

These two numbers are not supposed to match, and forcing them to agree is wasted effort. What helps is understanding how each one counts, then using whichever answers the question in front of you.

The main causes:

• Date assignment. Google Ads credits a conversion back to the date of the click, while GA4 records it on the date the event happened. A Monday click and a Friday purchase land on Monday in Google Ads and on Friday in GA4.
• Scope of credit. Google Ads reports conversions where a Google click or impression was involved. GA4 spreads credit across every channel that took part, so a user who clicks an ad and returns later through organic search can end up credited to that later channel in standard reports.
• Counting per click. Google Ads has a Count setting of One or Every. Set to One, three form submissions from a single click count once. GA4 counts every occurrence of the event.
• Time windows. The Google Ads click-through conversion window defaults to 30 days and can be set from 1 to 90 days, while the view-through window starts at 1 day and can go up to 30. Neither has to match the GA4 lookback window.
• Time zones. The Google Ads account and the GA4 property carry separate time zone settings. When they differ, daily totals stay permanently offset even though monthly totals sit close together.
• Cross-device and consent. Google Ads uses the click ID and signed-in user data to join devices, while GA4 needs Google signals or User ID configured. When a user declines measurement cookies, the two systems see different amounts of data.

A practical check is to compare monthly totals rather than daily ones and see whether the gap is stable. A gap that holds steady month after month is structural and explainable. A gap that jumps in a single week points at a tag change, a site change, or a goal configuration change made around that date.

For campaign goal structure and bidding setup inside the ad account, see the Google Ads management service.

How a conversion differs from a lead

A conversion is an event measured on your site or app. A lead is a real person with contact details who might become a customer. One conversion does not equal one lead, and in practice it usually does not.

The mismatches are ordinary. Someone unsure whether the form went through submits twice, which is two conversions and one lead. Someone who fills the form and then also phones in can be double counted in the sales system if records are not merged. A form filled by a bot is a conversion with no lead behind it at all.

The fix is to feed the downstream outcome back into measurement. When the sales team records in the CRM which leads closed, that data can be imported as offline conversions so bidding learns from lead quality rather than form volume. The question then shifts from how many forms arrived to how many customers were won and at what cost, which is the version management cares about.

For lead capture and handover between site and sales team, see the lead generation service.

Common mistakes and what each one costs

Most conversion measurement problems come from configuration nobody rechecked rather than from the tools. The recurring ones in hastily built accounts:

• Marking page_view on every page as a key event. The conversion count becomes the pageview count, and automated bidding chases people who open a lot of pages, which has no relationship to revenue.
• Double tagging, such as code embedded in the theme plus a second copy through Google Tag Manager. Totals double, reported CPA looks half of what it is, and a Target CPA set from that figure bids far too high.
• A thank you page reachable directly. Refreshes and bookmarks generate conversions that never happened.
• Not filtering internal and test traffic. On a low traffic site, a handful of test submissions distorts the whole month.
• Changing the conversion definition mid-month without an annotation, which makes month over month comparison meaningless and gets misread as a campaign effect.
• Measuring forms only when most buyers message through LINE, which understates the channels that drive chat and leads to cutting budget that was working.

A quick audit takes one pass: submit the form yourself from a phone, open DebugView to see how many times the event fires, then check the Google Ads account the next day to confirm the total moved by the amount you expected.

What is different about measuring conversions in the Thai market

Most conversion guides assume the transaction finishes on the website, which does not describe how many Thai buyers behave. The adjustments that matter:

Chat is where decisions actually land. Many users read the site and then move to LINE or Facebook Messenger. GA4 can capture the click out through outbound click tracking, but the conversation and the close happen outside the measurement system. The workaround is to count the click as a micro conversion, add source parameters to the link so the chat team knows which page sent the person, then import closed deals back as offline conversions.

Mixed payment methods make "order completed" ambiguous. A cash on delivery order or a bank transfer with a slip attached is not money received on the day the order was placed. Reporting all orders as one conversion makes the number look better than the bank account does. Splitting it into two events, order created and payment confirmed, is closer to the truth.

Traffic that ends on a marketplace is another case. When a user clicks from a brand site to a Shopee or Lazada storefront, the site owner sees only the outbound click and never the sale in their own analytics. Setting that expectation early prevents the misreading that the website generates no revenue.

Consent belongs in the picture too. Under the Personal Data Protection Act, a site collecting personal data needs a lawful basis and has to let users refuse. When users decline measurement cookies, some events disappear from reports. A drop that appears right after a consent banner goes live is usually a collection change rather than a sales decline.

Finally, Thai language pages. Forms with strict validation often reject Thai names, phone numbers typed with dashes, or addresses longer than the field allows. Users who hit a red error message tend to give up. Firing an event on form start as well as on successful submission turns that gap into a number instead of a guess. For rebuilding the pages that receive this traffic, see the landing page design service.

What to measure afterwards, and the honest limits

Once measurement is stable, only a few figures need watching alongside the rate: raw conversion counts, cost per conversion, the rate split by device and landing page, and the share of conversions the sales team confirms as real customers. Those four answer whether volume moved, whether it paid, where the problem sits, and whether the quality held.

The limit worth accepting is that no system captures everything. Users who decline cookies, users who switch devices mid-journey, browsers that restrict tracking, and people who see an ad then search your brand name three weeks later all produce gaps as a matter of course. Treat the numbers as comparable to your own history rather than as absolute truth.

The other limit is that measurement does not improve the rate by itself. It only tells you where to look first. The improvement still comes from the offer, the price, page speed, button clarity, and how easy the form is to finish.

Frequently asked questions about conversions

What is a conversion in one sentence

A conversion is an action on a website or app that the business has defined as valuable and configured the analytics system to count. Common examples are a form submission, a completed order, a call button tap, and a click through to chat. It does not automatically mean a sale, because it means whatever the person who set it up defined.

What conversion rate should a Thai business aim for

There is no single figure that answers this, because the rate depends on the conversion definition, the base unit, traffic mix, price point, and decision cycle length. The usable approach is to build a baseline from 8 to 12 weeks of your own history, split by device and channel, then measure changes against that.

Why does GA4 say key event instead of conversion

GA4 uses key event for important actions measured inside the property and keeps the word conversion for figures sent on to the advertising platforms. Separating the terms reduces confusion when the two sets of numbers disagree. The setup step is unchanged: switch on Mark as key event against the event in the Admin area.

Is it a problem that Google Ads shows more conversions than GA4

No, and it is expected given how differently they count. Google Ads ties conversions back to the click date and includes cross-device conversions GA4 may not see, while GA4 logs on the event date and may credit a later channel. The thing to watch is a gap that changes suddenly, not the existence of a gap.

How many conversion actions should one account have

Keep as few primary goals as the business questions require, usually one or two per sales path, and hold the remaining micro conversions as secondary events for analysis. Automated bidding pursues whatever is set as primary, so a mixed list of goals lets budget drift toward whichever result is easiest to produce, which is normally the least valuable one.

Defining conversions so they line up with money actually entering the business, then measuring them consistently, is work you do once and benefit from in every decision afterwards. If you would like a review of an existing setup or a fresh one built from scratch, the team is happy to talk it through.