Attribution

Attribution vs incrementality

Attribution is a credit rule. Last-click, first-click, linear, multi-touch, and platform data-driven each split the same conversion a different way. None of them equal incrementality (extra sales the ads actually caused). Marketing mix modeling asks a similar question at channel level, from weekly totals, without tracking a person.

Let's take a drink brand as an example. They sell a low-sugar sports drink for active women over 50, mostly from their own website. The drink brand buys YouTube, Instagram, and Google search. A 56-year-old runner sees a YouTube ad on Tuesday, an Instagram ad on Thursday, then Googles the drink brand on Saturday and buys a $36 case. Last-click attribution gives Google the whole $36. First-click attribution gives YouTube the whole $36. Linear attribution gives each channel $12. Three rules, one order, three stories.

Let's take a drink brand as an example. They sell a low-sugar sports drink for active women over 50, mostly from their own website. Meta (Facebook and Instagram) can also claim that Instagram-influenced $36 purchase. Google can claim the search. Add the two dashboards and the drink brand has invented extra revenue. That is why platform ROAS (return on ad spend: credited sales divided by ad spend) can look healthier than the bank account.

Why attribution matters

Budget meetings are attribution meetings. Let's take a drink brand as an example. They sell a low-sugar sports drink for active women over 50, mostly from their own website. If the drink brand lives in last-click attribution, YouTube looks expensive and Google looks like a hero. The YouTube ad may be the reason she searched. Agencies get stuck presenting six truths to one client.

The job of attribution is not to find the one true model. The job is to pick a rule, say it, and keep a second number (usually MER (marketing efficiency ratio: all company revenue divided by all ad spend) or a test) so you notice when the rule is lying.

How to read attribution

Pick an attribution model and write it on the slide. Last-click, first-click, linear, and platform data-driven will not match. Write the window too. A 1-day click and a 7-day click-plus-view are different products. Let's take a drink brand as an example. They sell a low-sugar sports drink for active women over 50, mostly from their own website. Changing the drink brand's credit window from 1 day to 7 days will make ROAS (return on ad spend) jump. That jump is a definition change, not a performance win.

Use MER (marketing efficiency ratio: total revenue divided by total ad spend) when the attribution argument needs to end: total drink sales over total ad spend, no model. For agencies, write the model into the reporting agreement. If the client lives in last-click and you live in Meta's data-driven model, every monthly will be a fight about whose truth is the slide.

Common attribution mistakes

• Adding up Meta ROAS (return on ad spend on Facebook and Instagram), Google ROAS, and TikTok ROAS and treating the total as company revenue. Let's take a drink brand as an example. They sell a low-sugar sports drink for active women over 50, mostly from their own website. The same case of drink can sit in all three dashboards.
• Changing the credit window (how many days after a click or view the platform still counts a sale) and calling it a performance win.
• Starving YouTube because last-click attribution never gives it the Saturday search. Let's take a drink brand as an example. They sell a low-sugar sports drink for active women over 50, mostly from their own website. The YouTube ad may be why she searched for the drink on Saturday.