WTF has Google Been Doing This Week? 15 SEP 2026
Google has just announced another big batch of updates designed to bring first-party data, measurement and AI much closer together. And while the headlines are about new features in Google Analytics, DV360 and Data Manager, there’s a much bigger message underneath it:
Google would quite like more signals to feed its AI
Google is now talking about measurement as a performance engine, rather than simply a report card telling you what happened after the money has been spent. In other words, Google wants advertisers to feed more measurement data and customer signals back into its systems to inform optimisation and improve performance over time.
In fact, Google says: “Every new signal you share improves your performance today and builds even stronger results over time.”
That sounds compelling. But not every signal is equally useful, and not every conversion deserves equal weight. Even if Google gets better at showing you what worked inside Google, it still can’t, by itself, tell you whether Google was the best place for that dollar.
The real trick is making sure the signals being shared help Google distinguish activity that looks valuable in-platform from activity that drives the business value that actually matters to you. That makes the choice of signals even more important. More feedback can help, but only if you’re feeding back the right things. Google isn’t going to solve that for you.
First-party data just became even more important
Update: Google is integrating Data Manager more deeply into Google Analytics and DV360, while also expanding enhanced conversions to help advertisers securely match customer data and improve ad relevance.
Google says advertisers connecting offline and app data to Data Manager see an average 26% increase in incremental ROAS.
AG Take:
That’s a significant claim, although as always with platform-reported benchmarks, we’d treat it as a reason to test rather than a result to expect. Google can provide useful evidence about what is happening inside its own ecosystem, but advertisers still need an independent view.
S**t in, s**t out (Harsh, but fair)
There’s an important catch. If the data you’re feeding the algorithm is incomplete, poorly structured, too broad, or not clearly weighted towards the outcomes that create real business value, giving Google more of it isn’t necessarily going to make things better. It could just help the machine get better at doing the wrong thing.
And this is where “more signals” needs some qualification. A lead isn’t necessarily a customer. A customer isn’t necessarily a profitable customer. And a conversion Google can easily observe isn’t necessarily the conversion you most want it chasing.
We’ve talked before about Negative Intelligence: teaching optimisation systems not just what success looks like but also what missteps look like, especially where apparent success fails to create commercial value.
Google’s latest push makes that feedback loop even more important.
Rather than simply feeding more to Google, the objective is to give the system better information about what deserves to be pursued and what deserves more weight. And (definitely don’t forget this part) what should be suppressed.
You already have useful data
Most businesses don’t have a perfectly organised data environment. That’s normal. (Phew, right?) But you have useful information sitting across CRM, sales system, ecommerce platform, call centre records, offline systems and spreadsheets that can provide signals about customer quality, profitability, sales outcomes and downstream value.
The job is to work out which of those signals are actually useful for media decision-making.
Sometimes that means better data integration. It might mean changing what Google is being asked to optimise towards once you connect sales outcomes back to earlier leads. Or it might simply mean using information you already have to make a better distinction between activity that looks successful and activity that actually creates value.
That is a much more manageable problem than “fix all our data”.
Measurement is getting more grown-up
Update: Google is pushing harder into the idea that no single measurement methodology can tell you the whole story.
Attribution tells you what happened. Incrementality helps tell you what actually caused it. Media mix modelling (MMM) helps you understand the broader contribution of your marketing investment.
Google is also putting more emphasis on Qualified Future Conversions, an approach announced earlier this year and detailed further this month. It’s designed to connect an early customer signal with conversions that happen much later, potentially up to 180 days after the original click.
AG Take:
That’s particularly interesting for businesses where the customer journey doesn’t end with a click or a lead form. Hello long conversion paths!
But better visibility doesn’t automatically mean better judgement. If Google can connect more downstream outcomes back to earlier activity, advertisers still need to decide which outcomes actually represent value.
And seeing more value attributed to Google still isn’t the same thing as proving Google created that value. That’s why attribution, incrementality and broader modelling need to work together.
AI is sitting in the middle of all of it
This is where it gets interesting.
Google has been working hard on automating the decisions that advertisers used to make manually: what to match against, who to prioritise, which creative to serve and where to send the user.
Update: Now it’s making the quality of the signals feeding those decisions a much bigger part of the equation.
Last week we talked about Google taking on more of the operational decisions. This is the other side of that shift.
AG Take:
Advertisers can still exert control. That control sits in deciding what represents value, what information the platform receives, what it should ignore or suppress, and what success actually means in business terms.
In order to do that, advertisers have to decide what to teach Google to optimise for.
Here’s what we’re checking in our clients’ accounts
1. What useful signals do we already have?
Before looking for more data, we’re looking at what clients already know about customers, sales, value, profitability and conversion quality, and which of those signals could meaningfully improve optimisation.
As our clients know, we’re always pushing for more of this business data to be brought into advertising and media strategy, and for media programs to have access to the information that helps distinguish a conversion from a commercially valuable outcome. That is how we tie advertising performance to profitability and prove it.
2. What is the algorithm being directed to optimise towards?
It bears repeating: A lead isn’t necessarily a customer. A customer isn’t necessarily a profitable customer. We’re checking whether the conversion signals being rewarded reflect actual business value, rather than simply the easiest actions for Google to see.
3. What should we stop rewarding?
This is where Negative Intelligence matters. We’re looking for evidence in the form of leads, customer types, conversion paths or behaviours that appear successful in-platform but fail to create enough value downstream.
4. Are enhanced conversions and other data signals configured correctly?
More data only helps if it’s being captured and matched properly.
5. Are we measuring incrementality, not just attribution?
If Google is reporting more conversions, we want to understand how much of that growth is genuinely incremental. And we don’t want the platform to be the only lens through which its own contribution is judged.
6. Are longer customer journeys being reflected?
For businesses with longer consideration cycles, we’re looking at whether the measurement framework captures value beyond the immediate click.
7. Where do we still need human guardrails?
The more decisions Google makes automatically, the more important exclusions, business rules and commercial constraints become. Google can optimise against the signals it receives. It can’t independently know everything that makes a customer valuable, unprofitable or simply wrong for the business.
8. Are we still looking at the whole media picture?
Because Google can tell us what happened inside Google. It can’t, by itself, tell us whether Google was the best place for that dollar.
So, what’s the opportunity?
Update: The direction is becoming pretty clear: more data, more signals, more automation.
AG Take:
“More” isn’t really the point.
The opportunity is to get better at deciding which signals matter, which outcomes deserve more weight, which should be suppressed and how to independently measure whether automation is creating real commercial value.
You don’t need perfect data to start doing that. You do need a clear view of what the business actually values. Of course you’re also going to have to figure out where that information lives, and how to turn it into better media decisions.
That’s where the work sits, and that’s the kind of work we’re doing.
