8 types of attribution every DTC brand should know.

Most brands install an AI-powered attribution tool and try to track every customer's journey perfectly. The numbers are precise and feel scientific. Start in the middle of the tree with a tool like that and you end up precise but inaccurate.

Start at the top with MER and nCAC instead. Add surveys and MMM for accuracy, then layer in MTA for precision within that accuracy, and you get both.

ROAS can mislead you. Brand A ran a 4.2x ROAS with a 35% new customer ratio, 48% COGS and 8% platform fees, and lost $15K a month. Brand B ran a 2.8x ROAS with a better new customer mix and tighter COGS, and made $45K in profit.

MER tells you if something is working. It can't tell you what, so you can't make channel decisions from it alone. It's the star on the tree, and you still need the rest of the tree.

Two formulas set the range you spend in. Everything else in this framework is about spending between them as well as you can.

Click-based attribution overvalues the lower funnel by 250% or more. We've seen top-of-funnel creative drive 13x more incremental acquisitions than bottom-of-funnel creative, while the click data told the opposite story.

Add the survey to your Shopify checkout with KnoCommerce, Fairing or a native option. List every channel you spend money on, such as Facebook and Instagram, TikTok, Google Search, YouTube, podcasts and friend or referral, plus “Other” with a text field. Keep it under 10 options and randomize the order.

Extrapolate the answers to all new customers, then work out your cost per new customer response in each channel. That tells you where to push budget.

People don't always remember correctly, and long purchase cycles make it harder. Some channels always under- or over-report, which is where MMM comes in to check the answers.

Surveys tell you what customers remember. MMM tells you what the data says is driving revenue, whatever people recall. Use MMM to check the surveys, and use survey data to calibrate MMM.

Its biggest benefit is marginal performance. You see where you are on the response curve as well as your average, which tells you where to put the next dollar.

When surveys and MMM agree, push budget hard. When they disagree, dig deeper before you make a big move.

You need $500K or more a year in ad spend, ideally two or more years of weekly data, clean history and someone who understands both the math and the business. If your data is messy or you plan to wing it, skip MMM, because it might do more harm than good.

It tells you whether retargeting is driving sales or only claiming credit, how much YouTube drives beyond clicks, whether cutting branded search by 50% would lose anything, and whether AppLovin or CTV is worth the investment.

The trade-off is time and money. Tests run 4 to 8 weeks and need geographic spread and enough sales volume to reach statistical significance. You're deliberately cutting spend in the test regions, so the opportunity cost is high. One well-designed test can still save you hundreds of thousands in spend that wasn't incremental.

With Facebook's dashboard you have to trust them. With MTA you can audit it, and first-click attribution shows you where someone originally discovered you.

Say someone finds you through UGC on Meta, sees a retargeting ad a week later, then Googles your brand name to buy. Last-click gives Google the sale. MTA shows you all three touches.

Plot CPA against spend for every ad in an account control chart. Your top ads should have a low CPA and high spend. If they don't, you have a media buying problem that no amount of new creative will fix.

MTA is still click-biased, and it overcounts total revenue across channels. Use it for ad-level and campaign-level decisions, and leave channel allocation to surveys and MMM.

Watch for view-through inflation. If a large share of conversions are view-through, the platform is padding your numbers, so force click-based optimization.

GA4 has its own trap. It defaults to last-click attribution, which heavily favors Google's own products. Someone sees your Meta ad, gets interested and Googles your brand to buy, and Google Search gets the credit for demand that Meta created.

Google built a free analytics platform that defaults to crediting Google. Use GA4 for web analytics like traffic and on-site behavior, and keep it out of the job of deciding what drove a sale.

That's why post-purchase surveys and MMM matter. They tell you what created the awareness, not what grabbed the last click.

Model 7 is the trunk. After iOS 14.5, pixels lost a lot of signal. CAPI or Elevar sends conversion events server to server, around the browser's restrictions. Without it your data leaks, and every other tool works with incomplete signal.

Model 8 is the roots, and it sets your targets. Return rate and purchase frequency drive most of the difference in LTV between channels.

Your cohort LTV tells you what CPA to target in each channel. That feeds the blended MER and nCAC target at the top, which feeds channel allocation in the middle, which feeds ad-level optimization at the bottom.

Skip this step and you optimize the whole tree toward the wrong target. You'll efficiently acquire cheap, low-LTV customers, which looks great on a dashboard and terrible on the P&L 12 months later.

In the example, Meta customers have a 35% return rate, 3.2x purchase frequency and a $180 12-month LTV. TikTok customers have a 20% return rate, 2.1x purchase frequency and a $95 LTV. Meta at a $40 CPA is a steal, and TikTok at a $30 CPA might be overpriced.

But every AI model I've seen starts with click-based data, then layers complexity on top of the bias. The result is a biased answer that looks more scientific.

The value of understanding attribution is knowing where your model is wrong, so you can compensate. I know surveys over-credit memorable channels, so I adjust. I know MTA is click-biased, so I adjust. AI attribution gives you a number without telling you where it's biased.

The right approach is top-down. AI tries to attribute every touchpoint perfectly and then roll up, and you can't build accurate channel insights from biased touchpoint data, however much AI you add.