The one-slide version
The whole approach on one canvas: three stages of maturity, each read through the three E's, standing on one data foundation.
The KPI tree — who sees which number
Most reporting fails by showing everyone everything. The discipline: three tiers, three audiences, and every metric knows which tier it belongs to.
The maturity ladder
The pattern in almost every audit: teams measure Efficiency obsessively, Effectiveness loosely, and Experimentation not at all. CPA can fall beautifully while the business gets nothing — you simply bought more of the customers who were already going to buy.
Basics — can you trust your own numbers?
“What happened?”
Descriptive truth: when a number moves, something real moved with it. Most failures I'm called in to fix are stage-one failures wearing a stage-three costume.
Conversion tracking tied to revenue
Events that map to money, carrying a value — not a count of form fills.
Core media metrics
The arithmetic layer — all computed on this site's free calculators.
CPC = spend ÷ clicks
CPM = (spend ÷ impr.) × 1,000
CPA = spend ÷ conversions
Break-even ROAS
A target derived from your margin, not guessed in a board meeting. Calculator.
Break-even ROAS = 1 ÷ margin %
MER & new-customer CAC
The whole-business view platform ROAS can't inflate; split new from returning.
New CAC = spend ÷ first-time buyers
Pacing & budget share
Catches underspent winners and last-week budget dumps.
Share of traffic
Your first competitive metric — position against a defined rival set.
Traffic quality & IVT
Cheap media that isn't human isn't cheap.
Creative & landing A/B tests
One variable, a pre-declared metric, enough sample to mean something.
Ready for stage 2 when
- Someone can explain the platform-vs-analytics gap on demand.
- Every revenue-relevant conversion is tracked — including off-site ones (WhatsApp, calls).
- Your break-even ROAS was derived, not assumed.
- One dashboard exists that finance and marketing both accept.
Intermediate — where is the value coming from?
“What contributed?”
Divides the outcome among the things that produced it, checks the media was actually seen, and measures your position in the market — not just inside your own account. Everything here is correlational.
Attribution studies
Rules-based, then data-driven — delete a channel from the modelled journey and see what it loses.
Time decay = 2−(t ÷ half-life)
Viewability
MRC standard: 50% of pixels, ≥1s display / ≥2s video — read against the measurement rate.
vCPM = (spend ÷ viewable) × 1,000
Attention & eye-tracking
Viewable means it could be seen; attention measures whether it was.
Cost per attentive second
Reach, frequency & saturation
Concentrated budget buys the same people repeatedly and calls it performance.
Incremental reach = Δ unique ÷ Δ spend
Share of voice (SOV)
Your presence relative to the category's advertising — the input side of growth.
Share of search (SOS)
Free, public, and a leading indicator of market share — see the charts below.
Share of market (SOM)
The outcome side — without it, ESOV can't be computed and budgets stay a negotiation.
AI answer-engine visibility
A growing share of buying starts with a question to ChatGPT, Gemini or Perplexity — track how often the answers cite you versus your competitor set, on a fixed prompt list.
Brand health tracking
Awareness, consideration, preference, NPS — on consistent waves so the trend is readable.
Cohort & LTV analysis
Turns a one-off CPA into a business case; payback is what finance wants.
Payback = CAC ÷ monthly contribution
Retail & commerce media closed-loop
Retailer networks tie ad exposure to actual sales data — the nearest thing to closed-loop ROAS, inside each retailer's walls.
Brand safety & suitability
A cheap CPM next to the wrong content is a brand cost no media report shows.
Platform conversion lift
A randomised holdout inside one platform — causal, within its own walls.
Ready for stage 3 when
- You can state each channel's contribution and defend the calculation.
- You know your viewability, attention and frequency — not just your CPM.
- You track share of search against a defined competitor set.
- Someone has noticed attributed revenue exceeds actual revenue — and it bothers them.
The share family — measuring the market, not just the account
The layer most performance teams never build. Three views of the same idea: your position, its trajectory, and what it predicts.
Efficiency metrics tell you how cheaply you bought. ESOV tells you whether you bought enough to grow. They are different questions — and only one of them predicts market share. Every share metric is only as honest as its denominator: define the competitor set once, write it down, never quietly change it to flatter a chart.
Brand health & mental availability
Performance measurement sees the demand you harvest. Brand measurement sees the demand you create — and the gap between them is where long-term value hides.
Category entry points (CEPs)
The occasions that make someone think of the category — creative builds the links, the tracker measures them.
Mental market share & penetration
Presence across all CEPs — correlates with sales share better than prompted awareness.
Distinctive brand assets
Colour, logo, sonic cue — measured on fame and uniqueness.
Advanced — what would have happened anyway?
“What caused it?”
The only stage that can take credit away from a channel. Everything below it can be gamed by claiming demand that already existed; these methods cannot.
Marketing mix modelling
Aggregate regression — no cookies, no consent needed. Wants 2–3 years of weekly data and real spend variation.
Saturation: Hill curve
Marginal ROI & saturation
A channel averaging 4× may return 1.2× on its last unit of spend — the number that should set the next budget.
Geo & matched-market tests
The cleanest causal evidence — power and minimum detectable effect settled before launch, not after.
DiD = Δtest − Δcontrol
Synthetic control
When you can't randomise: a weighted composite of untreated markets becomes the counterfactual.
Bayesian MMM, calibrated
Experiment results become priors that constrain the model — so MMM and geo tests stop disagreeing.
Brand lift studies
Randomised exposed-vs-holdout surveys — conversion-lift logic applied to memory instead of purchase.
Long-term vs short-term split
Activation shows in weeks; brand accumulates over years. Measure only the first and you under-invest in the second.
Price & promotion elasticity
Usually the largest term in an MMM — and it competes with media for credit.
Cross-media reach
Deduplicated reach across TV, online video and social — panels fused with platform big data.
Predictive CLV
Moves bidding from "who converts" to "who is worth acquiring".
The data foundation — what every method stands on
Every method above inherits its truth — or its garbage — from this layer: the measurement architecture I build, compressed to four blocks.
Collection & consent-era repair
Server-side tagging, consent mode, enhanced and modelled conversions — repairing what privacy removed from the observable layer.
Warehouse / composable CDP
One system of record for profiles, revenue and history — the single table every method reads from.
Data clean rooms
Privacy-safe joins with platforms and retailers — matched, aggregated, threshold-limited output only.
Governance & definitions
A metrics dictionary with named owners — the reason finance and marketing argue from the same number.
The operating rhythm — triangulation on three clocks
Mature measurement doesn't pick a winning method. It runs three on different clocks, each correcting the others, each reporting to a different room.
| Method | Decides | Cadence | Reviewed by | Bias |
|---|---|---|---|---|
| Attribution | What to change inside a channel | Daily / weekly | Team | Fast but biased |
| Share & brand tracking | Whether the market position is moving | Monthly | CMO | Directional |
| Incrementality experiments | Whether the other numbers tell the truth | In rotation | CMO | Accurate but narrow |
| Marketing mix modelling | How much budget each channel gets | Quarterly | Board | Unbiased but slow |
- Attribution Decides what to change inside a channel fast but biased
- Share & brand tracking Decides whether the market position is moving directional
- Incrementality experiments Decides whether the other numbers tell the truth accurate but narrow
- Marketing mix modelling Decides how much budget each channel gets unbiased but slow
When a geo test says a channel's true iROAS is 2.1 and attribution credits it with 5.0, you don't pick a side. Attribution keeps steering the week; the experiment's ratio becomes a correction factor whenever that number drives budget. Experiments are the ground truth that keeps the fast, biased methods honest.
The roadmap
Never buy a measurement method more advanced than your data hygiene. An MMM on unreliable inputs doesn't produce a rough answer — it produces a precise wrong one, and the precision is what makes it dangerous.
Common questions
What is a marketing measurement framework?
The agreed set of methods a business uses to answer three different questions: what happened, what contributed, and what actually caused the result. Dashboards answer the first, attribution and share metrics the second, and only experiments and MMM the third.
What should marketing report to the board?
Strategic outcomes, not channel diagnostics: growth attributable to marketing, market share and ESOV, MER, LTV:CAC, and the headlines from incrementality tests. CPCs and CTRs are diagnostic — they explain movement in those numbers and belong in the team's review, not the board pack.
What is share of search and why does it matter?
Your brand's share of category search volume. It acts as a leading indicator of market share — often moving months ahead of sales — and unlike share of voice it's free to calculate from public search data. That combination is why it has largely replaced SOV as a practical planning input.
What is ESOV?
Excess share of voice — your share of voice minus your share of market. Binet and Field's IPA work established that positive ESOV predicts share growth, at a long-run average of roughly 0.5pp of market share per year for every 10pp of ESOV. It is the rare metric that links spend to growth rather than to efficiency.
Attribution vs incrementality?
Attribution divides credit for conversions that already happened — correlational. Incrementality withholds advertising from a randomised control group and measures the difference — causal. A channel can dominate attributed conversions and produce almost no incremental revenue; branded search is the classic case.
Do I need marketing mix modelling?
Usually not until tracking is solid, media is genuinely cross-channel, and you have 2–3 years of consistent weekly history with real variation in spend. Below that, the same budget does more good fixing collection and running geo experiments.
What is mental availability?
From the Ehrenberg-Bass Institute: the probability a brand comes to mind in a buying situation. It's measured against category entry points rather than in isolation, producing mental market share, mental penetration and network size — and it predicts sales share better than prompted awareness.
How does privacy change measurement?
Consent, ad-blockers and walled gardens hide a growing share of the journey from user-level tracking. The answer is layered: server-side collection with modelled conversions repairs the observable layer; a warehouse becomes the system of record; clean rooms handle privacy-safe joins; and aggregate methods — MMM and geo experiments — need no user-level data at all. Privacy pressure is the main reason MMM returned to the centre of enterprise measurement.
How do I know which stage we're at?
Ask what your team would do if platform and analytics numbers diverged by 30% tomorrow. If nobody could explain the gap, you're at basics regardless of the tools you own. If you can explain it but can't say which channel earned the revenue, you're intermediate. If you can say what would have happened without a channel — and you've tested it — you're advanced.
Not sure which stage you're on?Send me last month's report. I'll tell you which questions it can and can't answer — and which rung to fix first.
Chat on WhatsAppMethods & terms on this page
- Marketing measurement
- KPI tree
- Board reporting
- Marketing mix modelling
- Bayesian MMM
- Incrementality testing
- Geo experiments
- Synthetic control
- Difference-in-differences
- Attribution modelling
- Markov removal effect
- Viewability
- MRC standard
- Attention measurement
- Eye-tracking
- Reach & frequency
- Share of voice
- Share of search
- Share of market
- Share of traffic
- ESOV
- AI answer-engine visibility
- Brand health tracking
- Mental availability
- Category entry points
- Distinctive brand assets
- Brand lift
- Adstock
- Saturation curves
- Marginal ROI
- Price elasticity
- Cross-media measurement
- Retail media measurement
- Data clean rooms
- Composable CDP
- Modelled conversions
- Consent mode
- Measurement governance
- Predictive CLV
- LTV : CAC
- Break-even ROAS
- MER
- Triangulation
Related: Marketing analytics consultant · Performance marketing consultant · Break-even ROAS calculator · Metrics glossary · Insights