How do you apportion TV spend to Amazon CAC?

Decide what share of TV spend Amazon gets, write the rule down, and add that share to Amazon's ad spend before you divide by new Amazon customers. The rest of TV goes to your DTC site. The rule can come from spend share, new-customer share, or an incrementality test, but everyone has to report on the same one.

If you just look at the dollars spent on Amazon, purely on sponsored ads, the CAC always looks dirt cheap. That isn't because Amazon is cheap. It's because TV is doing some of the work and getting none of the credit there. We built a CAC apportionment demo on a synthetic DTC brand so you can move the TV share yourself and watch both CACs change.

Key takeaways

• Amazon CAC built only on sponsored ads leaves out shared media, so it reads low whenever TV is driving Amazon searches.
• Splitting TV moves cost between channels: Amazon CAC goes up, DTC CAC goes down, and blended CAC stays the same.
• An incrementality test, such as a geo holdout, is the best evidence for the split. Spend share and new-customer share are rules, not proof.
• Pick one agreed-upon way of attributing, label it as the default, and keep one or two alternate model filters for comparison.
• Re-test the split when TV spend, creative, or the Amazon assortment changes enough to move the answer.

Quick answers

Why does Amazon CAC look so cheap?

Amazon CAC looks cheap because the spend side only holds sponsored ads, while the customer side holds everyone who bought, including people who saw a TV spot and then searched Amazon. The cost of the TV spot sits in the DTC row or in a brand line, so Amazon gets the customers and none of the cost.

TACoS has the same blind spot. TACoS is total advertising cost of sale: ad spend divided by total sales on Amazon, including organic sales. It is a useful Amazon efficiency measure, but if "ad spend" only means sponsored ads, TV is missing from it too.

If you've been living in a rolled-up world, spend sits in one pile, customers in another, and blended CAC is the only number everyone trusts. Blended CAC is still right for the whole business. It just can't tell you whether Amazon or your site is paying its way.

What counts as a new Amazon customer?

Use Amazon's new-to-brand counts as the customer side. Amazon Ads reports orders and sales from first-time customers of your brand on the Amazon store, within a 12-month look-back window. That window differs from how most DTC sites define a new customer, so write both definitions next to the numbers.

You can't match Amazon shoppers to site customers, so treat the two counts as separate channels rather than one deduplicated total. For mirrored orders, see how to avoid double-counting Shopify, Amazon and TikTok orders.

Which rules can you use to apportion TV spend?

There are four common rules for apportioning off media like TV. They differ in how much evidence sits behind them.

A geo holdout is the most practical incrementality test for TV. You go dark in matched regions and compare new customers on your site and Amazon against regions that kept TV on. Meta's open-source GeoLift package is one way to design and read this kind of test. The lift split between your site and Amazon becomes the share.

Check whether you can read Amazon new-to-brand results by region before you design the test. If you can't, test the site side and treat the Amazon share as an assumption.

A worked example: splitting 50% of TV to Amazon

These numbers are synthetic. A DTC brand spends $290,000 in a month and gets 4,500 new customers.

Blended CAC is $290,000 / 4,500 = $64.44.

In the rolled-up view, TV stays with DTC. DTC CAC is ($150,000 + $100,000) / 2,500 = $100.00. Amazon CAC is $40,000 / 2,000 = $20.00. That $20 is the dirt cheap number.

Now say the agreed view is that TV is impacting Amazon to the degree of half its effect. Amazon gets $50,000 of TV.

• Amazon CAC: ($40,000 + $50,000) / 2,000 = $45.00
• DTC CAC: ($150,000 + $50,000) / 2,500 = $80.00
• Check: $45 × 2,000 + $80 × 2,500 = $290,000, so blended CAC is still $64.44

Here is the same month under each rule. For the lift share, assume a geo holdout found 900 of the TV-driven new customers on Amazon and 600 on the site, so Amazon's share is 60%.

Against the rolled-up view, Amazon CAC rises by 53% to 150%. DTC CAC drops by $8.42 to $24.00. The total never changes, because apportioning moves cost between rows. It doesn't create or remove any.

TACoS moves the same way. If Amazon total sales were $300,000, sponsored-only TACoS is $40,000 / $300,000 = 13.3%. With $50,000 of TV added, it's $90,000 / $300,000 = 30.0%.

What does the split change in decisions?

The split changes which channel looks like it can take more budget. On the rolled-up view, Amazon at $20 looks like the place for every extra dollar. On the 50% view, Amazon at $45 and DTC at $80 are closer.

Compare each CAC with contribution, not revenue, since Amazon fees and repeat patterns differ. Contribution margin LTV by acquisition channel covers that side. For the next dollar rather than the average, use marginal CAC, as in blended, attributed, incremental, and marginal CAC.

How do you govern the agreed view?

Governance means everyone quotes the same Amazon CAC in the same meeting. We set it up in three parts.

One agreed view. Store the TV share as a dated setting in the warehouse, not a spreadsheet tab. Every channel CAC report reads that setting.

A couple of alternate model filters. Keep the rolled-up view and one test-based view as labelled options. They answer "what if we're wrong about the split" without anyone rebuilding the math.

A re-test cadence. Re-run the incrementality test when TV spend, creative, or the Amazon assortment changes enough to matter. Change the setting with a new effective date and leave history alone, so last quarter's numbers still tie out.

We define these as named metrics in a DTC semantic layer, so the rule lives in one place. If you're weighing an attribution tool against building this yourself, the Northbeam vs custom data warehouse comparison lays out the tradeoffs. Before any of this, check that spend and orders are clean with the profit dashboard readiness check.

Frequently asked questions

What to do next

Open the CAC apportionment demo and set the TV share to 0%, 50%, and your own best guess. Watch where Amazon and DTC CAC land, then compare that gap with what your team quotes today. If you want help setting up the agreed view on your own data, book a call with me.