Key takeaways
- Count as a conversion only the actions that sit on the path to money: a paid order, a signup, a lead your sales team accepts. Everything else is an engagement event, however useful it is.
- Send value, currency and a unique transaction ID with every purchase. GA4 deduplicates purchases that share a transaction ID, and an empty ID makes it merge every purchase into one.
- A conversion window decides how long after a click a conversion still counts. Google Ads defaults to 30 days and recommends at least 7; match the window to how long your buyers actually take.
- Ad platforms, GA4 and your order database will never agree exactly. Use the backend order count as the truth and compute how much each platform over-claims, then judge cost per acquisition on real orders.
- In AdPix, marking an event as a key event applies to past data immediately, revenue comes from the event's value parameter, and a stable gap of a few percent against GA4 is normal; a clean 2× or 3× gap is a bug.
Conversion tracking is recording the actions that make you money, such as a paid order, a signup or a lead your sales team accepts, and connecting each one to the visit and campaign that came before it. Done well, it gives you a cost per acquisition and a return on ad spend you can trust enough to move budget. Done badly, it counts thank-you page reloads, credits three ad platforms for the same order and reports a conversion rate that has nothing to do with revenue. This guide covers the four decisions that separate the two: what counts, how to stop double counting, which window to use, and how to reconcile the numbers when Google Ads, Meta and your analytics disagree.
Conversion tracking is the practice of recording business-value actions (orders, signups, qualified leads) as named events with a value, and attributing each one to the traffic source that produced it.
What should count as a conversion?
A conversion is an action that sits directly on the path to money, and most sites need only two to four of them. A shop counts paid orders. A SaaS product counts trial signups and paid subscriptions. A lead-gen business counts form submissions and, more importantly, leads that sales qualifies. Scrolling 75% of a page, watching a video or clicking a phone number are engagement events: worth tracking, wrong to optimize ads toward.
The test is simple. If the number doubled tomorrow, would revenue follow? If the answer is “probably not”, it is not a conversion. Google Analytics 4 calls these actions key events, and limits a standard property to 30 of them (50 on Analytics 360), which is a hint in itself (Google Analytics Help).
| Business | Primary conversion | Secondary conversion | Where it is confirmed |
|---|---|---|---|
| Online shop | purchase with value and order ID |
begin_checkout |
Payment callback on your server |
| SaaS | Paid subscription | Trial signup | Billing system |
| Lead generation | Qualified lead | Form submitted | CRM, days after the visit |
| Content site with ads | Newsletter signup | Returning reader | Email platform |
The last column matters more than it looks. A browser-side “thank you” page misses the buyer who pays and closes the tab before the gateway redirects back, and it fires again on every reload. The more money an action carries, the more reason to send it from the system that confirms it; the server-side tracking guide shows how, and why a tagging server is not the first step. The event tracking guide covers naming these events and sending CRM stages from the server. Saving the campaign on the order or lead record itself, as the first-party data guide describes, keeps each confirmed conversion tied to its source even after the browser has dropped the cookie.
How do you set up conversion tracking step by step?
Setting up conversion tracking takes seven steps, and only the fourth happens in the analytics tool’s admin screen. The rest is deciding and testing.
- List the conversions. Write the two to four actions from the table above that apply to you, with the system that confirms each.
- Name the events. Use lowercase names with underscores and Google’s recommended names where they fit (
purchase,sign_up,generate_lead). The same name must reach every tool. - Send value, currency and an ID. A purchase without
valueadds nothing to revenue. A purchase without a transaction ID cannot be deduplicated. - Mark them as key events. In GA4 you toggle “Mark as key event” or click the star next to an existing event. Marking affects reports from that point on, does not change historic data, and can take up to 24 hours to appear (Google Analytics Help).
- Choose the counting method. GA4 offers “Once per event”, the recommended setting, which counts every occurrence, and “Once per session”, a legacy setting that counts one per session (Google Analytics Help). Google Ads has its own setting: “Every” conversion suits sales, “One” conversion per ad click suits leads, where a second form from the same person is not a second customer (Google Ads Help).
- Tag every paid link. A conversion without a source is just a sale. Consistent UTM tags are what let the conversion land on the right campaign; the UTM parameters guide explains the fields, and the UTM naming convention gives a copy-paste standard plus a shared sheet that keeps every link on it.
- Test with a real transaction. Place a test order or submit a test lead, then find it in a debug view with its value, currency and ID before any campaign depends on it.
How do you stop double-counting conversions?
Double counting has three common causes, and each has a specific fix: a unique ID per conversion, one ID shared across browser and server copies, and one sending path per order.
Reloads and retries. Send a transaction ID (the order number) with every purchase. Google Analytics deduplicates purchase events that share a transaction ID (Google Analytics Help). The same page carries a warning that catches real stores: never send an empty string as the ID, because GA4 then treats every purchase with transaction_id="" as one and deduplicates them all. A missing or placeholder ID is worse than none.
The same event from browser and server. Many shops send a purchase through the Meta Pixel and again through the Conversions API, which is fine if Meta can match them. Meta deduplicates when the Pixel’s eventID matches the server’s event_id and the event names match, for events received within 48 hours of the first one with that ID (Meta for Developers). Generate the ID once, at order creation, and pass it to both.
Two paths in your own analytics. If your tool does not deduplicate across browser and server, sending one order down both doubles revenue. AdPix’s documentation says exactly this and recommends the browser for pre-purchase steps and the server for purchases, renewals and refunds (AdPix Docs).
A quick check: compare a week of analytics orders with the order database. A gap of a few percent is normal. Analytics showing more orders than the database almost always means duplicates.
What is a conversion window, and which one should you use?
A conversion window is how long after an ad interaction a conversion can still be credited to that ad. In Google Ads the click-through window defaults to 30 days and can be set from 1 to 90 days depending on the conversion source; Google recommends at least 7 days. Engaged-view conversions default to 3 days and view-through conversions to 1 day (Google Ads Help). Meta has its own attribution setting per ad set, so check it in Ads Manager rather than assuming it matches Google’s.
Pick the window from your buyers’ behaviour, not from the default. If most customers of a low-priced product buy within two days of the click, a 30-day window mainly adds credit for people who would have come back anyway. If you sell furniture or B2B services where the decision takes three weeks, a 7-day window hides most of what your ads produce, and Smart Bidding then optimizes toward the minority who buy fast.
To find the right length, look at the time from first visit to conversion for last quarter’s customers and set the window just past the point where about 90% of them have converted. Then keep it fixed. Changing windows mid-quarter changes every comparison you make against the previous period. The attribution models guide explains how the window and the model together decide which channel gets credit.
Why don’t Google Ads, Meta and analytics conversion numbers match?
They count different things on purpose. Google’s own page on conversions versus key events lists the causes: identity and consent settings, which channels can receive credit, the attribution model, the conversion window and key-event lookback window, the counting method, and whether reports use interaction time or conversion time (Google Analytics Help). Ad platforms also count conversions their ads touched, including view-through, while analytics usually gives each order to one session source.
| Reason | Ad platform (Google Ads, Meta) | Analytics (GA4, AdPix) |
|---|---|---|
| Who gets credit | Its own ads, any touch inside its window | One channel per order under the report’s model |
| Model | Google Ads conversions can use data-driven attribution | GA4 traffic acquisition uses the last non-direct session source |
| Timing | Can be reported by interaction time (the click date) | Reported on the day the conversion happened |
| Views | View-through and engaged-view can count | Clicks and visits only |
| Overlap | Each platform claims the same order | Each order counted once |
Worked example: reconciling one month (hypothetical)
Say a store spent 3,000 on Meta and 2,000 on Google Ads last month. Meta Ads Manager reports 120 purchases, Google Ads reports 90, and the order database shows 150 paid orders, 80 of which analytics attributes to those two paid channels.
- Platform claims: 120 + 90 = 210 purchases for 150 real orders, an over-claim ratio of 210 ÷ 150 = 1.4.
- Cost per acquisition by platform claims: 5,000 ÷ 210 ≈ 23.8.
- Cost per acquisition on analytics-attributed paid orders: 5,000 ÷ 80 = 62.5.
The real answer sits between those numbers, and the gap tells you how much of the platforms’ reporting is overlap and view-through credit. Track the over-claim ratio monthly. If it jumps from 1.4 to 2.0 without a change in spend, something changed in tracking or attribution settings, not in your customers. When you need to know what ads truly add beyond what would happen anyway, the incrementality testing guide covers holdout tests, and for the split of the whole budget, AdPix marketing mix modeling estimates how many conversions each paid channel caused rather than touched (Pro plan).
How does conversion tracking work in AdPix?
In AdPix a conversion is a key event: any event you mark under Admin → Website settings → Key events, typed exactly as it appears in the Events report. Adding or removing one needs the marketer role or above (AdPix Docs). Several details differ from GA4 and are worth knowing:
- The mark applies to history. AdPix reads the key-event list when a report opens, so marking an event counts all its past occurrences right away; removing the mark reverses that without rewriting data.
- Names match exactly.
Purchaseandpurchaseare different events, and a key event whose name matches nothing stays at zero with no warning. - Revenue comes from the event. The key-event revenue column sums the
valueparameter sent with each event; an event withoutvalueadds zero. - Purchases deduplicate by transaction ID. In the browser, a
purchasewith atransaction_idgets a stable ID, so reloads and retries collapse into one order (AdPix Docs). - Server events are idempotent. Orders sent through the server-to-server API carry an
event_id; re-sending the same ID and body writes no new event, and each order’s attribution is frozen as a last non-direct snapshot when it is first written (AdPix Docs).
Key events feed the Home conversion column, Traffic acquisition (key events, session key event rate, revenue), Explore and the funnel report. Funnels are on every plan including Free; the AdPix funnels guide shows how to build one from your key events. E-commerce reports, including sales and ROAS by channel, are also on every plan (AdPix pricing).
AdPix is a web analytics platform that replaces Google Analytics 4: realtime, acquisition, engagement, conversion and e-commerce reports, plus attribution, lift testing and marketing mix modelling, with no data sampling and a free plan. It runs alongside GA4, so you can track conversions in both and compare. For a team in Iran, where Google Ads is not offered to advertisers and GA4 opens only through a VPN, that second copy is more than a cross-check: the guide to Google Analytics risks in Iran explains why your conversion record should not live only in an account you do not control. Expect a stable gap of a few percent: AdPix does not model the behaviour of visitors who declined consent, while GA4’s consent mode can, and AdPix drops duplicate event IDs. A clean multiple, two or three times the GA4 number, points to a bug such as a tag installed twice (AdPix Docs). If your GA4 figure comes from an exploration over a long date range, check the data quality icon first: a sampled result is an estimate, and the GA4 data sampling guide shows when that happens. The guide to cookie consent and analytics shows how to size the consent part of that gap, and the website tracking guide puts conversions in the context of a full tracking plan. If you have not installed the tag yet, the AdPix install guide takes about five minutes.
Your next step: write down your two to four conversions with the system that confirms each, check that every purchase carries a non-empty order ID, and pull last month’s order count from your database so you can calculate your platforms’ over-claim ratio before you set next month’s budget.
ROAS calculator
Return on ad spend and your break-even point. The free plan covers 100,000 page views a month for each website, no credit card. Setup takes under five minutes.
Frequently asked questions
What is conversion tracking?
Conversion tracking is recording the actions that create business value, such as a paid order, a signup or a qualified lead, and tying each one to the visit and campaign that preceded it. It is what turns ad spend into a cost per acquisition and a return on ad spend you can compare across channels.
What is the difference between a key event and a conversion?
In Google Analytics 4 a key event is an event marked as important to the business, used for analyzing acquisition and behavior. A conversion is a Google Ads term for an action used for bidding and ad reporting, and it can be created from a GA4 key event. The two can show different numbers because they use different attribution, windows and counting settings.
Why does Google Ads show more conversions than Google Analytics?
Google's own help lists the reasons: different attribution models (Google Ads conversions can use data-driven attribution, while GA4's traffic acquisition reports credit the last non-direct session source), different conversion windows and lookback windows, different counting methods, consent and identity settings, and whether conversions are reported by interaction time or conversion time. Align these settings before treating a gap as a tracking error.
How do I stop duplicate conversions?
Give every conversion a unique ID and send it every time: a transaction ID on GA4 purchases, a matching event_id and event name when you send the same event through the Meta Pixel and the Conversions API, and never send one order from both the browser and the server unless the tool deduplicates across them.
Which conversion window should I use?
Use a window slightly longer than the time most of your buyers take from first ad click to purchase. Google Ads defaults to 30 days, allows 1 to 90 depending on the conversion source, and recommends at least 7 days; a short window undercounts slow purchases, a long one credits ads for sales they did not cause.
Sources and methodology
Google Analytics, Google Ads and Meta behaviour (key event limits, counting methods, transaction-ID deduplication, conversion windows, Pixel and Conversions API deduplication, reasons for discrepancies) comes from the help and developer pages in the sources, opened in November 2026. AdPix behaviour comes from docs.adpix.io pages opened the same month, and plan placement from adpix.io/pricing (read on 13 November 2026). The store figures in the reconciliation example are hypothetical.
- Google Analytics Help: About key events
- Google Analytics Help: Mark events as key events
- Google Analytics Help: Change the counting method of key events
- Google Analytics Help: Minimize duplicate key events with transaction IDs
- Google Analytics Help: Conversions vs. key events in Google Analytics
- Google Ads Help: About conversion counting options
- Google Ads Help: About conversion windows
- Meta for Developers: Handling duplicate Pixel and Conversions API events
- AdPix Docs: Key events
- AdPix Docs: E-commerce events
- AdPix Docs: Server-side tracking
- AdPix Docs: The numbers don't match GA4
This article was drafted with AI assistance and reviewed by the AdPix team before publishing. Editorial policy