Designing Marketing Measurement: Finding the Gaps in Your Company's Data with a Five-Layer Audit

For: marketers and team leads who work with marketing data  ·  Format: self-assessment workbook  ·  2026-07-11

All the Tools Are There, but There's No Picture

Open the GA4 report, check the Search Console data, dig through the conversion pixel logs, all three screens load just fine.

But faced with the question "is our marketing measurement actually working right now," not one of these screens can answer it alone.

Each screen only shows its own slice. GA4 only knows what happens after someone lands on the web, Search Console only knows the moment a search click happens, and the pixel only knows a single ad click.

Parts 1 through 9 of this series took those screens apart one at a time. Each post dug deep into what a cookie is, how to attach a UTM, how to read a GA4 report, and how to handle search volume and the dark funnel.

This finale does something a little different. Instead of adding one more new concept, it overlays the screens the earlier posts filled in onto a single diagnostic sheet.

This post keeps exactly one promise: one five-layer audit sheet.

But before you can overlay them, you first have to decide what shape to overlay them in, and that shape is the building in the next section.

Image: connecting scattered screens into one single sequence is this post's goal.

Measurement Is a Building: A Five-Layer Cross-Section

Thinking of marketing measurement as a single building makes it easier to grasp, because a signal originates on the top floor, gets filtered floor by floor on the way down, and is confirmed as revenue on the bottom floor.

The industry commonly calls this practice of planning layer by layer a measurement plan. Note upfront, though, that this isn't an official GA4 feature name, it's a practical concept shared by the Google Analytics Academy and various consultants.

Layer 1, Supply / Source Signal. When ads and content went out, how much was spent, how much exposure it got, all of that accumulates here.

Layer 2, Demand Expression. This refers to the movement of people who saw that signal, days later, searching the brand name or typing the site address directly.

Layer 3, Acquisition. The layer where that person actually opens a website session, and where they came from gets recorded at the browser level.

Layer 4, Conversion Ledger. That session turns into a concrete action, a lead form, an inquiry call, a signup, and gets tagged with identifiers like a UTM, click ID, or tracking number.

Layer 5, Confirmation. Whether that conversion actually turned into a signed contract or revenue gets finally confirmed.

Which of these five layers is already filled in and which is empty differs by company, and the fifteen questions in the next section pinpoint exactly where those gaps are.

Self-Assessment by Layer: Five Layers, Fifteen Questions

Answer three yes/no questions for each layer. Whichever layer comes back with two or more "no" answers, that's the gap in your company right now.

Image: of the five drawers, some are already lit, and some are still sitting in shadow.

If only one layer is empty, that's fortunate, but usually two or three layers are empty at once. Even so, don't try to fix everything at the same time. The prescriptions in the next section cover only the three most common combinations.

Two Ledgers: Don't Merge Them, Cross-Check Them

As you fill in the five layers, you're bound to run into one problem: the moment you force a tagged record and an untagged lump into a single table, a double-counting argument starts.

The solution is parallel tracking, not summing. Keep the ledger (record-level data with reliable tags), the estimate (untagged lumps allocated via time series), and the validation sample (a small sample like a self-report survey) separate, and just cross-check them every month.

An overseas case makes clear why a validation sample matters. In Refine Labs' 12-month study, the gap between automated attribution and self-reported responses widened to as much as 90%, and RevenueCat, on the basis of that gap, applied a 1.5x correction factor to a specific channel's reported performance.

Why these three ledgers are kept separate and how to build each one has already been covered in detail alongside the five-layer big picture in another post, so here we'll just bring in the conclusion. As established in The Two Ledgers, Five Layers Deep Dive: don't merge them, cross-check them.

Summary of the Person Key: A Three-Rung Ladder

There's another axis that runs across the five layers: the key that ties a person together. It climbs a three-rung ladder, from a browser key, to a login key, up to a universal key like a phone number.

A browser key lives only inside a single device. The moment a login happens, multiple browser keys get bundled under one person, and once a universal key, like a hashed phone number, attaches, web, calls, and offline all link to the same person.

Touchpoints that happened on a different device before login can't be restored in the ledger, and that's closer to a structural limit than a lack of technology. That share falls to the estimate ledger from the previous section to carry instead.

Why this ladder matters and what breaks at each rung is covered in more depth in the Dark Funnel Tracking Guide and The Two Ledgers, Five Layers Deep Dive.

Image: the browser key, login key, and universal key link together like a chain.

Prescriptions by Diagnosis: Three Common Types

Fill in the five-layer audit sheet, and the result tends to converge on one of three shapes.

Type 1, the Company with Only the Acquisition Layer

The case where you're only looking at GA4. Layer 3 is filled in, but Layer 1's supply signal and Layer 5's confirmed revenue are empty, so acquisition is clearly visible, but there's no way to know whether it ultimately turned into money.

The prescription has an order. Fill Layer 5 first. Just attaching a lead number or contract ID to revenue data you already have, so it becomes joinable, solves half the problem, and after that, start stacking ad spend and exposure as a time series in Layer 1.

Type 2, the Company Without a Ledger

The case where conversions happen but carry no tags. Web leads don't get a UTM attached, and phone inquiries all come in through a single main number.

The prescription starts with setting UTM rules, adding channel-specific tracking numbers one at a time, and saving click IDs. No matter how many untagged conversions there are, they stay as an estimate, never a ledger.

Type 3, the Company That Doesn't Stack Supply Signal

The case where ads keep running, but no record survives of when or how much was spent, or when content went out. Even if you want to cross-check against demand expression later, there's no counterpart to compare it to.

The prescription is simple but needs consistency. Write air dates and publish dates into a single event calendar, and pull a daily time series of ad spend by channel, that's enough.

The Monthly Cross-Check Routine: Turning Measurement into an Operation

Fill in the audit sheet once and stop there, and it's just a single snapshot. Turning it into an operation takes repetition.

On the same day every month, line up the ledger's channel split, the estimate ledger's allocation split, and the validation sample's response split.

If the three roughly agree, leave them as is. If any channel is out of line, put just that channel up for testing next month. If you're not sure, turning channels off and on in sequence works too, if it's a channel you're going to cut anyway, that's basically a free experiment.

Image: on the same day every month, put the three ledgers on the scale and compare.

Map of the Whole Series

Parts 1 through 9, it turns out, were all really about filling in some part of these five layers.

• Layer 1, Supply / Source Signal: Ad Effectiveness, Lift and Adstock Made Simple
• Layer 2, Demand Expression: Search Console & Naver DataLab, Dark Funnel Tracking Guide
• Layer 3, Acquisition: Cookies, Sessions, and Events Basics, GA4 Getting Started, 5 Reports, Getting Started with GTM
• Layer 4, Conversion Ledger: UTM Parameters for Beginners, Conversion Tracking, Pixel, Click ID, Server-Side
• Layer 5, Confirmation: no dedicated post, but The Two Ledgers, Five Layers Deep Dive covers how to connect Layer 4 and Layer 5.

Image: the scattered posts all converge into five layers in the end.

This map is the final page of this series, and drawing one line between the tools you already have comes before buying one more tool.

If you take away just one thing, let it be this.

Measurement isn't about buying more tools, it's about drawing one picture between the tools you already have.

A five-part deep dive digs into each layer of this audit. Off-site clicks are covered in tracking links and MMPs, exposure in measuring off-site exposure and its field record counting 380 SERP slots, channel label quality in UTM conventions and BigQuery audits, and event definitions in event taxonomy design.

Sources

• Measurement plan concept (not an official GA4 feature name): Boom Online, Google Analytics Academy Part 2
• KPI framework: Google, Think with Google KPI Framework
• Self-report vs. automated attribution gap: getrecast.com, HDYHAU Guide
• Self-reported attribution in practice: Outbrain, Self Reported Attribution Guide
• Limits of self-reporting: Ruler Analytics, Asking HDYHAU Isn't Enough

The five-layer, two-ledger structure in this post generalizes an actual measurement design document from a specific industry, with company names and detailed figures removed and reconstructed as an example. The 90% validation gap and 1.5x correction factor are measured figures from a published overseas case, and measured ratios for individual domestic industries are not included in this post.