I looked to cite this but couldn't find it on Google, so I may be mistaken...
We still largely don't know what half of advertising works and what half doesn't. There's just more metrics obfuscating that reality.
Source: worked on market research focused on identifying what advertising worked and what advertising didn't which Google white labeled
I recall a conversation with a CMO who was discussing how all their metrics indicated TV ads were a waste of money, so in a few of their DMAs they pulled TV ads and their sales in those areas dropped by double digits.
The additional metric fidelity is a false sense of security in the data. There's a lot more going on to drive someone to purchase than simply the ad tagging that gets picked up. It can be useful in optimizing spend a bit, but should be considered more like correlation than causation.
So you can add budget to things that are converting and see if that linearly leads to more conversions, just don't assume that it's as clear cut as a dashboard makes it out to be - it generally won't be and after a few low hanging fruit gains you'll hit diminishing returns because it's not actually as simple as "person sees banner ad A and clicks = 4% chance of sale."
For example, you might have 4% of people who clicked through your banner ad end up purchasing. But 80% of those 4% had previously read a CNET review for your product before seeing that ad. How are you getting that data point? And even if you do, it doesn't necessarily mean putting a link to that review on the landing page for the ad will result in much lift, as there may have been a selection bias where most people swayed by a CNET review were the ones who had already read reviews.
One of the most influential touch points is WoM. So certainly offline product recommendations aren't ending up in your dashboard, and exposure to social media positive sentiment isn't either.
Easily half to 2/3rd of the things someone sees that influence their purchasing decision aren't being picked up by an attribution model for conversion tracking.
And "old cable TV ads" . Still an audience of 72 million for those in 2023.
And cross platform attribution really isn't as straightforward as you seem to think.
Now, you run trial campaigns for your product, collect data, and calculate XX base on it, and transfer this XX into your bidding formula. Or trust Google/FB/etc bidding algorithms/ML to do this for you.
I'm glad your confidence in your methodology gives you peace of mind.
And this quote long predates the internet...
This also means there is a huge opportunity to disrupt google at this pont.
edit to add reference: https://en.wikipedia.org/wiki/Enshittification
I also think that disruption will come from something that looks different than google, maybe something like chatgpt, that doesn't appear to be a direct competitor at first but ends up taking a significant bite out of usage of google.
What they actually sell is a B2B SaaS for participating in auctions on those web properties.