Impressions hold, clicks fall, and branded search quietly rises. We rebuild measurement so you can tell the difference between losing visibility and losing the click — because those need completely different responses.
Most marketing measurement assumes a click precedes a conversion. When a buyer researches your category in an assistant, reads a summary that names you, and then types your brand into a browser three days later, last-click attribution records that as direct traffic and credits nothing.
So teams see non-branded clicks falling and conclude SEO is failing, while branded search rises and gets written off as 'brand doing well'. Those two numbers are frequently the same story, and the standard dashboard makes it invisible.
We rebuild measurement around that reality: model zero-click exposure explicitly, track branded lift as a proxy, add self-reported attribution at conversion, and report confidence honestly rather than assigning false precision to numbers that don't deserve it.
Full attribution rebuilds are expensive. Often a proxy layer gets you 80% of the insight for 20% of the effort.
Best for: Most teams. Start here and only go further if the decisions genuinely require more precision.
Best for: Enterprises with large budgets to allocate and the data maturity to support it.
Four streams. None of them alone tells the story; together they do.
Which of your queries trigger AI Overviews and how completely they answer, modelled against impression and click data to isolate the real click impact.
Branded query volume tracked against non-branded performance. Rising branded with flat non-branded is the clearest available signal of AI-influenced discovery.
A 'how did you hear about us' field at conversion. Crude, but it's the only place a buyer will actually tell you they asked ChatGPT.
Where you appear across a benchmarked prompt set, reported alongside traffic so visibility and sessions can be assessed separately.
Measurement design, implementation, and reporting that a human actually interprets.
Reviewing what your current setup can and can't see, where tracking is broken, and which decisions you're making on unreliable numbers.
Quantifying which queries lose clicks to AI Overviews and what that's actually costing, so prioritisation is evidence-based rather than anxious.
Fixing the configuration issues almost every account has — broken conversions, misattributed channels, unfiltered internal traffic, missing Search Console integration.
Implementing and analysing source questions at conversion, which frequently reveals AI influence nothing else can see.
A dashboard covering rankings, citations, traffic, and pipeline in one view, built to answer specific recurring questions rather than display everything.
A written analyst read on what moved and why, because a dashboard without interpretation is a number nobody acts on.
The blind spot differs by how you sell.
Self-serve signups attributed to direct traffic hide a research journey that started in an assistant. Self-reported attribution usually reveals this immediately.
Deals close months after the research that shaped the shortlist. Last-click attribution credits the final touch and misses everything decisive.
Product research increasingly happens in assistants and returns as direct or branded traffic, so paid gets credited for demand organic created.
Data split across regions, business units, and platforms makes a single honest picture genuinely hard to assemble.
Every engagement runs the RON Loop — diagnose, map, build, prove — then repeats it monthly.
We review what your current setup can actually see, what's misconfigured, and which decisions rest on unreliable numbers.
We model zero-click exposure and branded lift to identify where AI influence is happening invisibly.
GA4 and Console fixed, self-reported attribution deployed, and reporting built around your actual recurring questions.
A written analyst read each month on what moved, what caused it, and what to do — not a dashboard link.
The point isn't a prettier dashboard. It's being able to tell whether falling clicks mean falling visibility or a surviving visibility that stopped producing clicks — because one needs more SEO and the other needs a different conversion strategy.
The questions clients ask us most before starting. If yours isn't here, ask us directly on a consultation call.
Not cleanly, and we won't pretend otherwise. AI-assisted research usually returns as direct or branded traffic, so no analytics platform can trace it definitively.
What works is triangulation: branded search lift, self-reported attribution at conversion, and citation share movement. Together they're directional and genuinely decision-useful. Anyone selling precise AI revenue attribution is overstating what's possible.
Possibly, and it's measurable. If impressions hold steady while clicks fall on informational queries, that's the AI Overviews signature. If impressions fall too, it's a ranking or demand problem.
Separating those two is usually the single most valuable piece of analysis we do, because the responses are completely different.
A 'how did you hear about us' field at conversion, with a free-text option. It's crude and it works — a meaningful share of respondents will name an AI assistant directly.
It costs almost nothing and reveals things no amount of platform configuration can.
Almost never. Most teams have adequate tooling that's configured badly — broken conversion events, misattributed channels, no Search Console integration, internal traffic unfiltered.
We fix configuration before recommending new tooling, and usually there's no need for the second step.
With ranges and stated confidence. Where a number is solid we say so; where it's a modelled estimate we label it as one and explain the assumptions.
Reports that present modelled estimates as measured facts erode trust the first time someone checks.
Usually yes, and it's worth doing for B2B where the sales cycle is long. Connecting self-reported source data to closed-won revenue is where this gets genuinely persuasive internally.
It needs your CRM data to be reasonably clean, which is often the real constraint.
Audit and configuration fixes typically two to four weeks. Reporting build another two to three. Self-reported attribution needs a quarter of collection before the data is meaningful.
The interpretation layer is ongoing, since a report nobody reads monthly stops being useful quickly.
Measurement audit and reporting build is typically a fixed-fee project in the mid four to low five figures. Ongoing interpretation is included in retainers.
Detail is on our pricing page.
We'll model your real zero-click exposure, separate visibility loss from click loss, and show you where AI influence is happening invisibly.
No commitment · 45 minutes · Immediate value