The Marketing Measurement Playbook · 2026 edition

Your dashboard gives you a number.
This playbook gives you the truth.

Every ad platform claims credit for the same sale, so you end up making budget calls on numbers that can't all be right. That gap has a name: attribution debt, the 20 to 40% of ad budget that goes to channels taking credit instead of driving sales. This playbook shows you how to fix it: the basics you need, the advanced moves most teams skip, and the expert upgrades that tell you what actually caused a sale.

Benchmark
Trust in platform data
51%
of data leaders don't trust it
Benchmark
Touchpoints per sale
9.5
last-click credits just one
Illustrative
Hidden value / brand
$329K
credited to the wrong channel
Benchmark
Causal vs last-click gap
9.5%
of revenue credit moves
The tour: how the three tiers fit together ≈ 2 min

The problem has a name: attribution debt

It usually surfaces on a Tuesday. The board deck says Meta drove 38% of revenue. Someone asks why the store disagrees with the dashboard by $50K on a single day. Nobody in the room can answer.

Your dashboards disagree because they're each counting the same sale. The fix isn't a prettier dashboard. It's knowing which spend actually caused the revenue.

Add up what every platform says it converted and you'll get well past 100% of your revenue. All that overlap is attribution debt: budget spent on a story, paid again every month. This playbook takes you up the ladder, from honest basics to the question every expert team ends up asking: what happens to revenue if I switch this channel off?

Attribution by channel

Illustrative
Causal Last-click
Same spend, two models. Where the credit really lands.
Paid social24% · 12%
Influencer / organic18% · 6%
Paid search15% · 22%
Direct / branded12% · 31%
Last-click gives too much credit to the bottom of the funnel (direct, branded search) and too little to the social and creator spend that created the demand. The playbook teaches you to see this gap. Tier 03 teaches you to measure it.

Price your debt

What does attribution debt cost you?

For brands running on platform-reported numbers, 20 to 40% of ad spend typically goes to the wrong place. Drag the slider to your monthly ad spend and see what that means in money.

Monthly ad spend$50K
Illustrative Estimated annual attribution debt $120K – $240K ad money going to the wrong place each year, at the 20 to 40% benchmark. Not all of it is recoverable. None of it shows up on a last-click dashboard.
The $99 analysis costs less than 5 hours of that debt.

How to use this playbook

Three tiers, meant to be run in order. The advanced and expert moves only work once the basics under them are clean. Each tier has a short video walkthrough and plays you can put to work this quarter.

01 · Basics

The groundwork every store owner needs. Do these and your numbers stop lying to you.

02 · Advanced

Move from "who got credit" to "what actually caused it": surveys, holdout tests, cohorts.

03 · Expert

Measure what would have happened without each channel, continuously and with confidence ranges. This is where a data team, or Causality Engine, earns its keep.

TIER 01 Basics · Fundamentals

Get your foundation honest.

Before any clever model, four things have to be true: you know the real economics of an order, you trust one revenue number, your tracking is clean, and you judge the business on blended reality, not platform-reported ROAS. Get these right and half of your "attribution problem" disappears.

PLAY 01.1Economics

Know your real unit economics

You can't divide up a budget you can't price. Most store owners know their AOV and blended ROAS, and almost nothing about what an order actually earns after all its costs.

The move
  • Build a one-line P&L per order: revenue − COGS − shipping − fees − ad cost.
  • Track contribution margin % and CAC payback in days, not "ROAS."
  • Set a floor: the blended return your margin can actually survive.
Watch Contribution margin % · CAC payback (days)
PLAY 01.2Hygiene

One tracking convention, enforced

Messy UTMs and duplicate pixels are why your reports fight each other. Fix the input and everything downstream gets more honest for free.

The move
  • Adopt one UTM naming convention and lock it in a shared sheet.
  • Move to server-side tagging to recover the signal cookies drop.
  • Audit your "(direct)/(none)" bucket. A big unknown share is a leak, not a channel.
Watch % clean-source sessions · unknown %
PLAY 01.3Cadence

A weekly numbers ritual

Measurement that isn't a habit decays. A short, fixed weekly review beats a beautiful dashboard nobody opens.

The move
  • One page, same day each week: MER, nCAC, CM%, spend by channel.
  • Record one decision the numbers drove, or admit you made none.
  • Flag the biggest week-over-week swing and ask "why," not "who."
Watch Decisions-per-week · MER drift
PLAY 01.4★ Highest leverage

Judge the business on blended MER

Platform ROAS is graded by the platform. Blended MER, total revenue divided by total ad spend, is the one number nobody can inflate for you. If you only run one play from this tier, run this one.

The move
  • Track MER (blended ROAS) and new-customer CAC as your two lead numbers.
  • Stop making spend calls on a single channel's in-platform ROAS.
  • Reconcile revenue to your store. The store is the ledger, not the ad account.
Watch MER · new-customer CAC (nCAC)
Building an honest one-page weekly review, live ≈ 3 min
TIER 02 Advanced · The needs you're skipping

Separate correlation from cause.

Clean basics tell you what happened. They don't tell you what caused it. This is where most brands get stuck: still trusting last-click while branded search and retargeting quietly take credit for demand you already paid to create. These four plays are the manual ways to get at cause.

PLAY 02.1Reframe

Ask the right question

"Which touch got the sale?" is the wrong question. "What happens to revenue if I switch this off?" is the only one that changes a budget.

The move
  • For every channel, write down what revenue would be with it and without it.
  • Treat branded search & retargeting as suspects, not heroes.
  • Rank spend by the return on the next dollar, not total credited revenue.
Watch Marginal ROAS by channel
PLAY 02.2Experiment

Run a geo or holdout test

The gold standard you can run yourself. Turn a channel off in some regions, keep it on in matched ones, and compare what happens to sales.

The move
  • Pick matched regions; pause one channel in test, keep it live in control.
  • Measure the lift in actual sales, not credited conversions.
  • Turn the result into a ratio you can plan budgets with.
Watch Sales lift % vs. control regions
PLAY 02.3Cohorts

Judge channels on payback, not the first order

A channel that looks expensive on day one can be your best channel by month three. Judge it on when customers pay back, not on the first order.

The move
  • Cohort new customers by acquisition month and channel.
  • Track cumulative contribution margin to the payback line.
  • Fund channels by LTV:CAC and payback window, not launch-day ROAS.
Watch Payback window · repeat rate · LTV:CAC
PLAY 02.4★ Highest leverage

Add a post-purchase "how'd you hear" survey

The cheapest truth signal you're not collecting. Live by tomorrow, paying for itself within a month. Asking customers directly catches the shares, group chats and word of mouth no pixel ever sees.

The move
  • One question at checkout: "How did you first hear about us?"
  • Compare what customers say to what your platforms claim.
  • Watch the difference. That gap is your attribution debt, in numbers.
Watch Self-reported vs. platform-claimed mix
The catch

These work, but they don't scale. A holdout test measures one channel, in one window, and takes weeks. Surveys drift. By the time you have one answer, your spend, your creative and the market have all moved. Real budget decisions need the answer for every channel at once, refreshed continuously. That's the wall Tier 03 breaks.

Don't want to spend a quarter at the wall? The $99 analysis is the shortcut. One upload, answers the same afternoon. See your gap for $99 →
Designing a geo holdout that survives scrutiny ≈ 3 min
TIER 03 Expert · The upgrades

Know what would have happened anyway.

This is the frontier: working out every channel's real contribution at the same time, pricing the next dollar before you spend it, and putting a confidence range on every recommendation. Done by hand, it takes a data scientist and weeks per refresh. This is the layer we built Causality Engine to automate. You should understand the moves either way.

PLAY 03.1Causal

Give credit for cause, not clicks

Assign credit by what a channel actually added, not by which touch came last. A model that knows what would have happened anyway stops you paying full price for demand you already had.

The move
  • Replace last-click rules with an estimate of what each channel actually added.
  • Subtract the baseline: demand a channel merely intercepted.
  • Report the causal number next to the platform-claimed one, and mind the gap.
Watch Causal vs. last-click contribution
PLAY 03.2MMM

Find where each channel stops paying back

Every channel has diminishing returns. The question isn't "is Meta good?" It's "at what spend level does the next dollar on Meta stop paying back?"

The move
  • Chart return against spend per channel and find where the next dollar stops paying.
  • Account for the lag: spend today drives sales next week.
  • Plan budgets on the next dollar's return, not the average.
Watch Marginal ROAS at current spend · saturation point
PLAY 03.3Confidence

Report ranges, not guesses

A number without a confidence range is a guess with good posture. Expert teams report ranges, like "Meta drove between 34 and 42% of revenue, with 95% confidence", and re-check them weekly.

The move
  • Put a confidence range on every budget recommendation.
  • Keep a rolling calendar of holdout tests to keep the model honest.
  • Raise spend only where the return is proven and clears your margin floor.
Watch Confidence % · model drift vs. tests
PLAY 03.4★ Highest leverage

What-if reallocation, before you spend

The payoff of everything above: try moving $X from channel A to B and see the projected revenue change before committing a dollar of budget. This is the moment measurement stops describing the past and starts pricing the future.

The move
  • Test each budget move in the model before you make it, not in last month's report.
  • Tune the mix to your margin floor, not a vanity ROAS target.
  • Ship the winning scenario, then compare what happened to what was predicted.
Watch Predicted vs. realised incremental revenue
Where this goes

This is the tier a spreadsheet can't reach. Working out every channel's would-have-happened-anyway number, refreshing it weekly, and carrying confidence ranges is a full-time data-science job. Causality Engine is the answer to your attribution debt: it runs all of this on your own store and analytics data. The Attribution Mismatch, What-If Simulator and Hidden Value views are these four plays, automated. Under the hood it's Bayesian causal inference, the same family of methods as Google's Meridian, so when your CFO asks why does the model say this?, you can actually answer. The playbook is yours to run by hand. When you want it live in an afternoon instead of a quarter, that's us.

Reading a What-If reallocation on real data ≈ 3 min

Self-diagnostic

Where are you on the ladder?

Tick what's true today. Be honest, nobody's watching. Your tier and your next move appear below.

Your tier

Tick the statements above to see where you sit, and which play to run next.

Brands that did this

The plays in this book run real stores.

Me Gorgeous The Two Sisters Twinkels OFFFTRACK Maison Tanger
Read the case studies →
34%
of ad budget reclaimed in 60 days, across Shopify and Magento stores.
OFFFTRACK · from the case studies

Where the free playbook ends

You have the map. The terrain is your data.

Run these twelve plays and you'll be measuring better than most brands your size. But a playbook can only describe the moves. It can't tell you your attribution debt, or where your next dollar pays back hardest. That answer only exists inside your numbers.

We kept this deliberately practical and deliberately incomplete: the manual versions of Tier 03 take a data team and weeks per refresh. If you'd rather see your own gap in an afternoon, with the confidence ranges to defend it, that's exactly what the $99 analysis is for.

See your own gap · $99, one-time

One upload. See what your dashboards can't.

Bring your analytics export and store revenue. We'll show you the gap between what your platforms claim and what actually drove your sales, the hidden value you're leaving on the table, and how certain we are about it. $99 tells you what $5,000-a-month tools tell their clients.

One-time, no subscription You own the report forever No contracts, nothing to cancel First report before lunch
1 Upload your paths CSV
2 See your Attribution Mismatch
3 Reallocate with the What-If Simulator

Trusted where it counts: 216+ brands run their weekly causal check on Causality Engine.