Tracking and attribution basics

0:00 Tracking and attribution

Here's a puzzle that confuses almost every new marketer. Your store recorded a thousand orders last month. Meta says it drove five hundred and twenty. TikTok says three hundred and ten. Google says two hundred and forty. Add them up and the ad platforms claim more than all your sales. And your analytics tool says only seven hundred and sixty purchases happened at all. Is everything broken? No. In this lecture you'll learn how tracking works, how attribution shares credit, why every tool gives a different number, and how to use each number for the right job. By the end, you'll know which number to trust for which decision.

0:47 The measurement toolkit

Let's start with what tracking actually does. Every time someone visits your site, clicks an ad or opens an email, tools record events. The toolkit has five parts. Web analytics, like Google Analytics 4, records sessions, sources and conversions on your site. Platform pixels and server-side APIs send conversion events back to ad platforms so they can report and optimise. UTM parameters are tags on your links that say where a visit came from. A tag manager, like Google Tag Manager, manages all those tracking codes. And your store or CRM holds the real orders, leads and revenue. That last one is your source of truth.

1:33 Attribution models

Now attribution. Customers often touch several things before buying: a TikTok video, a Meta retargeting ad, a Google search. Attribution decides how to share the credit. Last click gives it all to the final click, which over-credits branded search and retargeting. First click gives it all to the first touch, which over-credits discovery. Linear splits it equally. Position-based favours the first and last. And data-driven uses an algorithm to estimate each touch's contribution. In GA4, data-driven is the default, and the only alternatives now offered are last-click variants. Think of it like a football goal. Who gets credit: the striker, the midfielder who passed, or the whole team?

2:20 Why numbers disagree

Here's why numbers disagree. Each ad platform uses its own attribution window, for example counting a purchase within seven days of a click or one day after someone viewed an ad. And each platform only sees its own ads. So a shopper in Jeddah who watched a TikTok creator video, then clicked a Meta ad, then searched your brand on Google and bought, might be claimed by all three. One sale, three claims. Meanwhile GA4 misses some sales entirely, because some visitors decline cookies, use blockers or switch devices. The key idea: platforms over-claim together, analytics under-counts, and your store is the only complete record.

3:06 Example 1: one sale, three claims

Let's make it concrete with a simple example. A shopper in Jeddah sees your abaya in a TikTok creator video on Monday. On Wednesday a Meta ad reminds her and she clicks, but doesn't buy. On Friday she searches your brand on Google and buys. TikTok counts a view-through conversion. Meta counts a click-through conversion within seven days. Google counts the search click. GA4, using data-driven attribution, might split the credit across the channels it saw. So which is right? They all are, by their own rules. The mistake is adding them up and treating the total as real sales.

3:49 Example 2: monthly reconciliation (illustrative)

Now the realistic version, a monthly reconciliation. The numbers are illustrative. A Jeddah fashion store lines up every source against its store orders. Store: one thousand orders, that's the source of truth. GA4: seven hundred and sixty purchases, or seventy-six percent, because of consent refusals, blockers and cross-device journeys. Meta: five hundred and twenty, fifty-two percent, on a seven-day click and one-day view setting. TikTok: three hundred and ten. Google Ads: two hundred and forty. Sum of the ad platforms: one thousand and seventy, or a hundred and seven percent of real orders. Now they have ratios, and ratios are what you watch over time.

4:35 Watch the ratios

Why ratios? Because a sudden change is a clue. If next month Meta jumps from fifty-two percent of store orders to eighty percent, that's almost never a miracle. It's usually a duplicated pixel, a changed conversion setting or a tracking bug. Here's another trap, especially in cash on delivery markets. A Lahore electronics store counted a purchase when the checkout form was submitted, but around a quarter of those orders were refused at the door. The ad platforms were learning from orders that never paid. The fix was to count revenue only on delivered and paid orders, and send a separate delivered event back to the platforms.

5:22 Right number, right job

So which number do you use for what? Use your store or CRM for business decisions: revenue, profit, new customers. Use each platform's own reporting to compare ads and audiences within that platform, because the rules are consistent inside it. And use GA4 for on-site behaviour and cross-channel trends. Then, because click-based attribution keeps losing visibility, add methods that measure real impact. Incrementality or lift tests compare people who saw ads with a similar group who didn't. Geo tests switch spend on in some cities and off in others. And marketing mix models, like Google's Meridian and Meta's Robyn, use aggregated data over time.

6:07 Mistakes and what's next

A few common mistakes. Trusting platform-reported ROAS without comparing it to actual sales. Tagging internal links on your own site with UTMs, which overwrites the real source. Changing conversion definitions in the middle of a campaign. And judging awareness campaigns on last-click sales, which will always make them look useless. And a quick look ahead. Server-side conversion APIs, Google's enhanced conversions and modelled conversions under consent mode are the modern toolkit. They're covered hands-on in the Privacy-First Measurement course, and using them to steer ad automation is the focus of AI Performance Marketing.

6:47 Healthy measurement

How do you know your measurement is healthy? Three signs. The reconciliation ratios are stable month to month, and you can explain them. Every campaign link carries UTMs, and every key conversion is tested with a real transaction after any site change. And budget decisions reference store data and at least one blended metric, like MER or new-customer CAC, not a single platform's ROAS. Think of it as a simple measurement checklist: conversions firing correctly, pixel and server events deduplicated, a documented UTM convention, one agreed source of truth, and a fixed review rhythm.

7:28 Recap and try this now

Let's recap. Tracking records events; attribution shares the credit; and every tool counts with its own rules. Platforms over-claim together, analytics under-counts, and your store is the source of truth. Watch the ratios between them, fix sudden jumps, and in cash on delivery markets count only delivered and paid orders. Use each number for its own job, and add lift tests or mix models for true impact. Here's your try this now. Build the reconciliation table from the lesson text for last month: your store orders, GA4, and each ad platform, with each one's share. Write one sentence explaining the biggest gap.