Most attribution debates treat measurement error as one problem with one direction. Your dashboard is either too generous to a channel or too harsh on it. Pick a better model, the thinking goes, and the number gets more honest.
This week's Paid Traffic Truth broke that assumption in a single row. YouTube showed up distorted in two opposite directions at once, inside the same account, in the same week. Last click made it look slightly worse than it really is on revenue. Blended CAC made it look dramatically better than it really is on new customer cost. Neither distortion is visible if you only run one report, and they do not cancel out. They compound.
If you have ever defended or killed a channel off a single dashboard number, this is the post that explains why that number was never enough.
Here is the YouTube picture from Issue 009, week of Aug 23 to 29.
On revenue credit, YouTube ran a first click ROAS of 0.50 and a last click ROAS of 0.44. The channel opens more relationships than it closes, so last click hands some of its earned credit to whatever search or email touch came later. Read only the last click report and YouTube looks like a slightly weaker revenue driver than it is.
On acquisition cost, the distortion flips. YouTube's blended CAC read $319. Its verified first-time customer cost was $826, a 159% markup. Only 39% of YouTube's tracked buyers that week were genuinely new, so the blended number was mostly cheap repeat purchases wearing a prospecting costume. Read only the blended CAC and YouTube looks far cheaper at winning new customers than it is.
Same channel. Same week. One report understates its value, the other overstates its efficiency. A marketer running last click alone trims YouTube for looking weak. A marketer running blended CAC alone scales YouTube for looking cheap. Both are wrong, and they are wrong in opposite directions.
They come from two unrelated mechanics, which is exactly why no single model catches both.
The revenue distortion is a timing problem. Attribution windows decide who gets credit based on where a touch sits in the journey. Channels that open the journey lose credit to channels that close it. This is why every video and social channel in the grid, Meta, YouTube, TikTok and Pinterest, showed negative credit gaps on last click, while the search channels that close deals, Google and Microsoft, showed positive ones.
The cost distortion is a denominator problem. Blended CAC divides spend by every customer acquired, new and repeat together. Repeat buyers convert cheaply, so the more a channel leans on people who already bought, the more its blended cost gets dragged down. Meta at 87% new barely moves, a 14% markup. YouTube at 39% new explodes to 159%.
One error is about which touch gets the credit. The other is about which customers get counted. They live in different reports, run off different math, and have no reason to point the same way. Fixing one does nothing for the other.
The practical failure is not that your numbers are slightly off. It is that the two errors push your judgment in contradictory directions and you cannot see it happening.
Consider the same channel graded three ways:
Blended CAC says: YouTube costs $319 to acquire a customer. Looks mid-pack. Hold or scale.
Last click ROAS says: YouTube returns 0.44. Looks like a laggard. Trim it.
The verified view says: YouTube costs $826 for a genuinely new customer, but that customer nearly triples in value to $244 within a year. This is not a cheap channel or a weak channel. It is a payback-window channel you judge on months, not on week-one math.
Only the third read supports a real decision. It tells you YouTube is expensive up front, real on the back end, and evaluated on a timeline. Scale it if your cash flow can carry the payback window. Chill it if it cannot. Neither the blended number nor the last click number gets you to that call, because each one is answering a question you did not ask.
You do not need a fourth report. You need to stop trusting any single one and verify at the level the platforms cannot grade themselves.
Not modeled, not surveyed. Match against first party order IDs so you know the real share of new buyers behind any blended number. This kills the denominator distortion.
The gap between them is the signal. A negative gap flags a channel that opens relationships and is losing earned credit. This kills the timing distortion.
A high nCAC is only a problem if the customer never grows into it. YouTube's does. TikTok's, at $758 with far shallower back-end growth, is a different conversation.
Meta at $88 nCAC and 87% new is a genuinely different decision than YouTube at $826 and 39% new, even though blended CAC prints them at $77 and $319.
The through line from this week's Paid Traffic Truth is that every platform grades its own homework, and it grades in whatever direction flatters it. Blended CAC flatters on cost. Last click flatters the closers and shortchanges the openers. The only way to stop grading a channel two contradictory ways at once is to verify new versus repeat yourself, at the order level, and price the decision off that.