The short answer is: There is no universal marketing evaluation method. Any magic formula in a super-coach presentation is inevitably a lie, because companies differ in market, transaction cycle, margin, channel, product stage, and data availability. strict system of principles and toolsfrom which they are assembled his The following is what «always works» is what you do. not It works, and how to build a measurement system that makes real management decisions, rather than drawing beautiful dashboards.

What exactly are we evaluating?

Marketing efficiency is the ability of marketing actions Create incremental (additional) profits considering time-lapse, risk and capitalIn a normal company, measurement is built on three planes:

Financial performance
Incremental revenue → gross profit Contribution to margin/EBITDA/FCF After all the variable marketing costs.

EBITDA – profit before interest, taxes, depreciation and amortization; FCF – free cash flow.
Behavior and funnel
Attractions, activations, retention, frequency, average check, repeat purchases, referrals – all measured by cohort level.
Causality (causality)
Than. precisely Is it the way things change (channels, creativity, prices, stocks, seasonality)? It’s the field of experimentation, attribution, and models.

Without these three layers, any «method» is cosmetics.

Popular “methods” – in fact

ROMI/ROI/ROAS – beautiful interest rates, dangerous solutions

ROMI (Return on Marketing Investment) and ROAS (Return on Ad Spend) — useful locally.

ROMI – ROI – ROI – ROI – ROI – ROI – income on advertising expenditure.

The main sin: consider “income per RUB costs” without checking incrementalismSome of the sales would have happened without the baseline, but the numbers are happily attributed to advertising.
Revenue swaps. ROAS=500% sounds cool, until you think of 15% margin, 8% returns and discounts.
ignor lagIn B2B/long cycle, the deal closes months later; ROAS is anemic today, tomorrow is hypertrophied.
Cannibalization and “halo” effects are not accounted for: the promo on SKU A ate up SKU B sales and worsened overall margin contribution.

Conclusion: ROMI/ROAS without incrementality and margin idle metrics in business suit.

B2B – Business for Business; SKU – Individual Stock Keeping Unit.

LTV/CAC is the king on paper.

LTV/CAC (lifetime value per cost of attraction) is a chic ratio, if you count rightlyWhich is rare.

LTV – Lifetime Value; CAC – Cost of Customer Acquisition Cost.

LTVs often take «hospital average,» without cohorts, discounting and churn-models.
CAC forget to clean organics and brand: in the numerator only pay-in.
Marketing is not responsible for all LTV: product, support, logistics contribute the lion’s share.

Conclusion: LTV/CAC — strategy-compassWe need cohorts, discounts, outflow scenarios.

NPS is not money, it is mood.

NPS It is convenient and often useful, but:

NPS is the Net Promoter Score (Net Promoter Score) index.

money-linkage sluggish;
The response culture and sample bias distort the picture.
To sink the budget for the sake of +5 p.p. NPS with a negative unit economy is a managerial error.

Conclusion: NPS — alarmNot the meter.

Last-click/first-click – accounting, not attribution

The rules “last click”, “first click”, “linear” – technical heuristics, not a causal assessment.

skewed in favor of the bottom of the funnel or brand queries;
see view-through and cross-device;
In the world of cookie restrictions and walled gardens, it is simple break down.

Conclusion: They are suitable for management, but not for budgetary decisions.

* View-through – the effect of browsing without clicking; cookie – the ID file in the browser; walled garden – the “closed ecosystem” of the advertising platform.

MTA (multi-touch attribution) – beautiful on slides, painful in life

Yeah, MTA’s better than rules.

It depends on detailed tracking, which is becoming less and less (see below).GDPR/ATT/sandboxes);
Overlapped ecosystems (Meta, Google, TikTok) break the end-to-end ligament.
The result is unstable and vulnerable to modeling “for the desired answer.”

Conclusion: MTA is a useful lens, but not the only.

MTA: Multi-Touch Attribution; GDPR: European Data Regulation; ATT: Apple App Tracking Transparency Policy.

MMM (marketing mix modeling) is a big hammer.

Econometrics on aggregated data (weeks/regions) basic demand, seasonality, lags, saturationThat’s it. strong instrument, however:

It requires a lot of clean data and discipline.
It gives you medium elasticities. I don’t see the creatives and the audience. at the user level;
introduction is long, retraining is frequent, collinearity is insidious.

Conclusion: MMM — strategic calculatorBut it has to be crossed with experiments.

MMM – Marketing Mix Modeling (MIM)

Brand lift/exposomers – noisy and expensive

Measurements of “knowledge raising” by surveys/control groups are useful brandbut

The effects are small, the confidence intervals are wide.
transfer long-chain assumptions.

Conclusion: apply ROMI.

A/B Tests: Gold in Only Skillful Hands

Experiments are the main instrument of causality, but:

typical sins: «peeping», lack of power, cross-section (interference), seasonality;
B2B and Expensive Purchases Cycle Makes Tests long-term.

Conclusion: A/B — must have, if There is enough traffic, time and discipline.

A/B test is a controlled experiment with two variants (A and B) comparing metrics.

“Share of voice”, “market share”, “coverage” – long, expensive, unmanageable

It’s good for the Big Game, but:

market share. delayed and a rough indicator;
SOV It does not guarantee profits in a negative unit economy.

Conclusion: snooping, yes. budget — Careful.

SOV (Share of Voice) is a share of voice.

What is universal (and why is it not a “method”)?

Incrementalism
How much sales/profits add Your actions are beyond the baseline trajectory.
Ideal tools: experiments (geo-split, holdout), quasi-experiments, MMM.
Marginality
Not the revenue, but the revenue. margining: margin profit = gross profit — variable marketing costs — discounts / retours.
Time horizons
Tactics (0-90 days): performance, conversions, ROAS with adjustments.
Operations (quarter-year): LTV/CAC, retention, cohort economy.
Strategy (year+): market share, brand metrics, price per attention share.
Layers of measurement
Financial → Behavioral → Causal → Model → Operating.
Together they give a picture, separately — distortion.
Cost of measurement
Measure. preciselyhow much is it pays off.

Geo-split/holdout – geographical experiment with control regions; performance – performance marketing.

Working framework of the methodology: «5-layer measurement system»

(1) Financial layer (P)&L-contribution

KPI: Marginal Profit After Marketing (CMP), EBITDA, FCF.
Tools: Plan-fact on campaigns / channels / SKU, accounting for discounts / retours, accrual lags.
Rule: Any marketing report ends cash-in.

* P&L is the profit and loss statement; KPI is the key performance indicators; CMP is the contribution margin after marketing.

(2) Behavioral layer (funnel and cohort)

KPI: CAC, conversion by stages, Retention/Repeat Rate, AOV, frequency, churn.
Tools: cohort tables, RFM, user paths.
Rule: mean values — evil; solutions on cohort dynamics.

AOV – average check (Average Order Value); RFM – segmentation by prescription / frequency / amount (Recency, Frequency, Monetary); churn – outflow; Retention – retention.

(3) Causal layer (incrementality)

KPI: uplift, IROAS (incremental ROAS), delta-profit.
Tools: A/B, geo-experiments, holdouts, PSA tests, non-useful blackouts.
Rule: At least 10-20% of the time spent control groups.

* IROAS — incremental ROAS; PSA test — a test with a «pseudoreclam» / social announcement for calibrating the effect.

(4) Model layer (generalization and forecast)

KPI: channel elasticity, saturation, cross effects.
Tools: MMM (Bayes/Frequency), light-MMM on weeks, causal time series simulation.
Rule: model non-replace experiments, and calibrate on them.

Light-MMM – lightweight model of marketing mix; Bayes / frequency – approaches to statistical inference.

(5) Operational layer (management and discipline)

KPI: SLA data coverage UTM/call tracking, the share of «not set», the reporting speed.
Tools: Match tables, event hygiene, a single nomenclature of campaigns.
Rule: garbage at the entrance — garbage at the exitWithout it, any MMM would be a fortune teller.

SLA – service target level (data quality/timeline); UTM – traffic markup tags (Urchin Tracking Module).

“Recipes” for different businesses (skeleton that can be built)

E-commerce/D2C with fast cycle

North-star: contribution to margin/order (post-promo), percentage of incremental orders.
Tactics: A/B on promos and creatives, geo-holdout on major channels, MMM-light Quarterly.
Hygiene: Retours, promo-cannibalization, saturation curves.

D2C – direct sales to the consumer (Direct-to-Consumer); North-star – the main target metric; post-promo – calculation “after the promotion”; MMM-light – lightweight MMM.

Offline retail/O2O

North-star: Incremental revenue in test regions.
Tactics: Geo-experiments, cash data, mixed-media with TV/OOH, «spillover» modeling.
Hygiene: Brand search as a dependent variable, weather/calendar control.

O2O – online to offline; OOH – out-of-home; spillover – “flow” effect to neighboring regions / audiences.

B2B/long cycle

North-star: pipeline contribution (marginal contribution) SQL→Won).
Tactics: MQA/MQLMulti-link lead-account-campaign, lift tests of “clusters” of accounts, attribution at the level of capabilities.
Hygiene: De-duplication of leads, quality of source, SLA sales.

* SQL – Sales Qualified Lead; Won – Deal Winned Status; MQA – Marketing Qualified Account; MQL – Marketing Qualified Lead.

Mobile Apps (ATT/Privacy)

North-star: pLTV-D30/CAC by cohorts and IROAS.
Tactics: Creative tests on grids, network level incrementality, MMM for top fly.
Hygiene: SKAN-Windows, probabilistic attribution, post-Buck validation.

* ATT — App Tracking Transparency; pLTV-D30 — predictive LTV by day 30; SKAN — StoreKit Ad Network (attribution to Apple).

Clear formulas (and how not to spoil them)

Marginal contribution of the campaign
Contribution = (Revenue × Gross Margin) − (Variable Marketing Cost + Discounts + Returns)

ROMI (on profit)
ROMI = (Incremental Profit − Marketing Cost) / Marketing Cost
Key. incrementalWithout experiment/model, it’s divination.

ROMI – Return on marketing investments.

ROAS (honest, post-promo)
pROAS = Incremental Gross Profit / Ad Spend
Count. after refunds/discounts.

ROAS – income on advertising expenses; pROAS – post-promo ROAS (after corrections).

CAC (clean)
CAC = (Paid Incremental Spend on Acquisition) / (# New Paying Customers from Paid)
The organic and brand components are sideways.

CAC – Cost of attracting a customer.

LTV (cohort, discounted)
LTV = Σ_t (ARPU_t × Gross Margin × Retention_t) / (1 + r)^t
Where r is the discount rate, t is the months/weeks.

LTV – lifetime value of the customer; ARPU – average revenue per user (Average Revenue Per User).

IROAS (incremental ROAS)
IROAS = (Sales_with − Sales_without) × Margin / Ad Spend

IROAS is an incremental ROAS.

Why the “universal method” is a myth

The business model decides. The same ROAS in subscription and in single sale means different.
The data is different. Somewhere there are box office rows for 5 years and traffic is a million / day, somewhere — 200 transactions / month.
Legal environment breaks the trekking (GDPR/ATT).
Goals are different. Market share gain RUB Operating profit here and now.

GDPR is the European Data Regulation; ATT is App Tracking Transparency.

Frequent anti-patterns (to avoid getting stuck)

Optimization by average ROAS underfunding growth.
Budget as percentage of revenue Disciplined stagnation.
“Last click decides everything” → zeroing the top flyd and brand.
“All on brand / organic” → vulnerability to competitors, burnout base.
Fetish on NPS/Likes/Graps minus the money, plus the slides..
Micro-KPI without P communication&Local victories, global defeats.
Reports without lags/retours → drawn profits.

KPI – Key Performance Indicators; P&L is the income statement; ROAS is the revenue on advertising expenses; NPS is the consumer loyalty index.

How to build your own method: a practical plan (90 days)

Weeks 1-2. Diagnosis.

Describe the value path: source → lead → activation → purchase → repetition.
Fix P&L through channels: margin, discounts, returns, variable costs.

* P&L is the income statement.

Weeks 3-4: Layers and goals

Record North-Star and «guardrails» metrics.
Implement basic cohort reporting (min. two cohorts: paid / organic).

North-Star is the main target metric.

Weeks 5-8. Causality

Run 1-2 geo-holdout tests on major channels.
Inside the channels are system A/B tests of creatives and landings.
Implement non-useful “switch-offs” on tail campaigns.

Geo-holdout – Geo-holdout experiment with control; A/B test – experiment with two variants.

Weeks 9-12: Models and regulations

Build light-MMM (weeks x regions) with calibration according to experiments.
Adopt rules: UTM Charter, Event Codebook, Not Set Control < 5%.
Synchronize finance and marketing: a single nomenclature, accounting lags.

Light-MMM – lightweight model of marketing mix; UTM – tags for marking up traffic.

Next, the control circuit.

Monthly Budget Committee: decisions only on layers 1-3 + sanity check models.
Quarterly revisions of elasticities and saturation curves.

How to understand that the method is “live”, not a presentation

Decisions have changed. You’re really redistributing budgets, closing down ineffective campaigns and see The effect is money.
Transparent lags and retours. There are no magic “plusses” at the end of the month.
Convergence of sources. Experiments, MMM and Attribution not matchbut they are understandably divergent.
The cost of measurement is under control. Do not build an LHC to count CTR buttons.

LHC – jokingly: “Large Hadron Collider” as a metaphor for excess complexity; CTR – click-Through Rate.

Summary for the supervisor

There is no single method and there can be no. There. frame The principles (incrementality, margin, horizons, layers) on which your system is assembled.
ROMI/ROAS/LTV/CAC/NPS/attribution/MMM — toolEveryone works in their own zone and breaks down outside.
The main KPI of marketing – contribution to cash flowIt is proven causally and supported by data discipline.
The method has to change the decisions. If the reports are beautiful, and the budget lives out of habit — there is no methodology.

* KPI – Key Performance Indicators; ROMI/ROAS/LTV/CAC/NPS/MMM – see the transcripts above.