Foundations · Lesson 4 of 6

Tracking and attribution basics: pixels, postbacks and last click

About 11 minutes · Yogender Kumar, Managing Partner

By the end of this lesson you'll be able to

  • Explain how a pixel and a postback record a sale
  • Spot when several platforms are claiming the same order
  • Choose a source of truth and a deduplication rule for paying partners

Why this matters

Every pricing model in Lesson 2 depends on one thing: counting results correctly. If two partners each claim the same sale, you pay twice. If tracking misses sales, good partners go unpaid and stop working with you.

The core idea

When someone clicks a partner's link, the link carries a click ID, a unique reference for that visit. The job of tracking is to connect that click ID to an order, if one follows.

There are two main ways to do it.

A pixel is a small piece of code on your order confirmation page. It runs in the shopper's browser and tells the partner a sale happened. It's quick to set up. But ad blockers, privacy settings and cookies that have been cleared can all stop it, so some sales never get reported. In the UK, cookie consent banners cut what a pixel sees further still.

A postback, also called server-to-server tracking, doesn't rely on the browser to report the sale. Your system confirms the order, then your server sends the click ID straight to the partner's server. It takes more set-up, but it's far more reliable, and you can send it only once an order is confirmed.

Attribution is the next question: when several channels touched a sale, which one gets the credit? The most common rule is last click. The last channel the customer clicked before buying takes all of it.

Two ways a sale gets reported

Pixel (browser)

Shopper reaches the thank-you page
A tag in their browser fires
Partner records the sale

Postback (server to server)

Your system confirms the order
Your server sends the click ID
Partner records the sale
A pixel relies on the shopper's browser. A postback goes from your server to the partner's, so ad blockers and cleared cookies don't stop it.

Worked example

Here's one illustrative Northfield customer:

  • Day 1: clicks a social media ad, browses, leaves.
  • Day 3: reads a review site run by an affiliate partner and clicks through. Leaves again.
  • Day 5: searches "Northfield Home", clicks the paid search ad and buys a £60 order.

The social platform counts a sale, because its ad was clicked within seven days. The affiliate network counts a sale, because the review site's click was within its 30-day window. The search platform counts a sale, because its ad was the last click. Three reports, one order.

Across a month it adds up. The platforms report 120, 200 and 150 sales: 470 between them. Northfield's own system took 350 orders. That's 120 more sales claimed than Northfield actually took.

The cost is real. Northfield pays affiliates £15 per sale. Suppose 50 of the network's 200 sales also had a later paid search click. Paying for those 50 costs £750 for sales the search budget had already bought.

Northfield's sales in one month: claimed and real

Social platform reports

120

Search platform reports

150

Affiliate network reports

200

Added together

470

Orders the shop actually took

350
Three platforms claim 470 sales between them. The shop took 350 orders. The extra 120 are the same sales counted more than once.

Choosing a rule

There's no perfect attribution model. What matters is that you pick one, write it down and apply it the same way every month. Last click is easy to understand, which is why partners accept it. But it gives nothing to the channels that introduced the customer. Track 7 covers testing what each channel really adds.

The rule that saves the most money is simple: your own order system is the source of truth. Every sale you pay for must match a real order number.

In practice

  • Use postbacks wherever a partner supports them, and fire them only on confirmed orders.
  • Pass the order number with every conversion, so you can match it to your own records.
  • Agree a deduplication rule, such as paying an affiliate only when it was the last paid click.
  • Set a validation period, often 30 to 60 days, so returns and cancellations come off before you pay.
  • Each month, compare the total of all platform reports with your own order count.

Common mistakes

  1. Adding up every platform's dashboard. Each one counts the sales it touched. The total will be higher than your real sales.
  2. Paying on pending sales. Pay once the validation period has passed, after returns and cancellations have come off.
  3. Changing the rules mid-month. Partners plan around your rules. Agree changes in advance and put dates on them.

Try it yourself

In one month, search reports 90 sales, social reports 60 and your affiliate network reports 110. Your shop recorded 200 orders. By how many sales do the reports exceed real orders, as a number and a percentage? You pay affiliates £12 per sale, but matching against your orders shows they were the last paid click on only 80. How much would paying all 110 overpay?

Show answer

The reports add up to 90 + 60 + 110 = 260. That's 60 more than the 200 real orders, or 30% too many. Paying 110 sales instead of 80 overpays by 30 × £12 = £360 this month.

Quick check

Not scored. Pick an answer to see the explanation.

  1. What is a postback?

  2. Why do commissions usually stay pending for a month or two before they're paid?

Key takeaways

  • A click ID links a visit to an order. Pixels report it from the browser, postbacks from your server.
  • Every platform counts the sales it touched, so their totals overlap.
  • Your own order system is the source of truth for paying partners.
  • Agree a deduplication rule and a validation period before you go live.