Summary: Higher Meta ROAS can coincide with lower profit. Start with confirmed revenue, contribution and new-customer costs. The ten metrics in this guide then help explain measurement, creatives and delivery. A universal target cannot replace your own cost calculation.
Meta assigns purchases under its attribution rules. The shop confirms paid orders. Finance accounts for VAT, returns, product costs, fees and subsidised delivery. These views answer different questions and can each be internally consistent.
A budget decision needs them together. Use comparable periods, currencies and countries, and label each data source. The ten metrics below are organised into three levels: business and measurement, creative diagnosis, and delivery and customer mix. Reviewed on 28 September 2026.
Level 1: business and measurement
1. MER: overall marketing efficiency
MER = confirmed total revenue / total marketing costs.
State the denominator explicitly. Here it includes media, agency and creative production. Some reports use advertising spend alone; that different definition should not silently enter the same time series.
Illustrative calculation: €60,000 confirmed revenue, €10,000 media and €2,000 agency/production give MER 5. Excluding the additional costs produces 6; this guide calls that second calculation blended ROAS. All amounts exclude VAT, with revenue after returns and discounts.
MER is not profit. Show contribution after product costs, payment fees and other variable expenses alongside it, and account for marketing costs. Do not subtract discounts twice if net revenue already includes them. Falling MER alongside rising platform ROAS justifies investigation; it does not prove a particular cause.
2. New-customer CPA and complete CAC
New-customer CPA = defined media spend / confirmed new buyers.
A fuller CAC can also include allocated agency, production or sales costs. Document that allocation; media CPA and complete CAC are not interchangeable.
If €1,000 media produces 100 orders, order CPA is €10. If only 60 buyers are new, new-customer CPA is €16.67, approximately 67% higher. This is arithmetic, not an industry benchmark. Useful limits depend on margin, returns and demonstrated repeat-purchase value, rather than a general fashion or beauty table.
The shop or CRM determines new-customer status from purchase history. GA4 counts new users through signals such as first_visit or first_open; a new visitor is not a confirmed first-time buyer. Ads Manager does not provide a general personally identifiable buyer list for this reconciliation. Google explains its user metrics.
3. Event Match Quality as technical diagnosis
EMQ describes how customer parameters support matching events to Meta accounts. Its score is neither a percentage of recovered purchases nor a financial KPI. Even a high score does not establish accurate value, deduplication or permitted processing.
For each important event, check source, timing, value, currency, shared identifiers and permitted matching fields. Then monitor EMQ and errors in Events Manager. Meta’s CAPI training addresses matching alongside deduplication and data freshness.
In our ongoing engagement with Czech online retailer Erkado, work combines server-side event tracking, segmented campaigns across three distinct goals (e-commerce sales, B2B leads, and showroom hiring), and mobile checkout UX adjustments. This demonstrates that reliable tracking infrastructure establishes data clarity, but must be paired with clear campaign objective segmentation.
4. Separate blended ROAS from platform ROAS
Blended ROAS = confirmed total revenue / total media spend.
Meta ROAS = revenue attributed by Meta / Meta advertising spend.
The second formula depends on the attribution conditions selected. The first includes overall revenue, including other sources of demand. It therefore does not automatically remove organic sales or prove advertising’s incremental effect. MER differs here because its cost denominator is broader.
A ROAS target must fit your contribution calculation. A value of 4 can be viable at high margin and inadequate with expensive returns. Neither a universal minimum nor a general correction for “inflated Meta attribution” is defensible without relevant data.
Level 2: creative diagnosis
5. Hook rate: response to the opening
Hook rate = 3-second video plays / impressions × 100.
This is an explicitly defined analytical ratio, not evidence of purchase intent. Compare like placements, audiences, periods and suitable video lengths; very short videos have different conditions. A high ratio can show attention without additional orders.
Test understandable openings: the product in use, a specific problem, an outcome with supporting context or a customer question. Start by changing the opening while keeping the rest comparable. A universal 20%, 30% or 40% target cannot determine creative quality or whether to scale.
6. Hold rate: attention after the opening
Hold rate = ThruPlays / 3-second video plays × 100.
Record the ThruPlay column used and its current account definition. For longer videos, the 15-second viewing measure provides a reference; for shorter videos, completion contributes differently to the comparison. Do not mix those lengths without explaining the difference. Hook and hold are analytical conventions, not official profitability targets in their own right.
A strong opening followed by weaker retention can justify testing pace, explanation and proof. It does not automatically establish clickbait. Check clicks and confirmed purchases too. With few plays, a seemingly large ratio difference provides limited evidence.
7. Outbound CTR: clicks leaving the platform
Outbound CTR = outbound clicks / impressions × 100.
This differs from CTR across all interactions. A like or profile visit is not an outbound website click. An outbound click also does not establish a loaded page or a purchase.
Read outbound CTR alongside landing-page arrivals, loading speed and conversion at each funnel stage. Strong click response with weak sales can reflect audience, product, price, shipping or checkout. A consistent period and click definition matter more than a universal 1% or 3% target. Calculate cost per outbound click using the same click denominator.
Level 3: delivery and customer mix
8. CPM by placement
CPM = media spend / impressions × 1,000.
Feed, Stories, Reels and available Audience Network placements can have different prices and results. Include confirmed purchases, new buyers and contribution alongside CPM. Cheap reach alone does not justify moving budget.
Calculation on the same basis: 10,000 Reels impressions at €4 CPM cost €40. Ten confirmed purchases give €4 CPA. 10,000 Feed impressions at €12 CPM cost €120. Thirty purchases also give €4 CPA. Three times as many purchases do not make Feed more efficient here: cost and quantity increase in the same proportion.
Before excluding placements, check volume, conversion quality and potential effects on the whole campaign. A controlled test can be useful. The example does not establish general DACH CPM inflation or require switching to manual placements.
9. Frequency and reach
Frequency = impressions / people reached in the selected period.
The period is part of the definition. Weekly frequency is not directly comparable to monthly frequency. The average also does not describe how impressions are distributed between individuals.
Rising frequency together with deteriorating results can suggest constrained reach or creative fatigue. Check spend, reach, audience overlap and creative changes. A value above 2.5, 5 or 7 does not automatically require switching an ad off; remarketing and acquisition have different conditions.
10. ROAS for new and returning buyers
Separate confirmed first and repeat purchases using order history. Calculate cohort returns only with a documented allocation of revenue and spend. When campaigns serve both groups, total media cost cannot be split between them without a justified method.
High returning-customer ROAS can reflect valuable demand or attribution of purchases that were already likely. Lower acquisition ROAS can be viable over time if repeat purchases and margin are demonstrated. An invented lifetime-value assumption does not make losses sustainable.
Attribution asks which contacts receive credit for an order. Incrementality asks which orders would not have occurred without advertising. With enough data, holdout or geographic tests can help. Small samples, differences between countries and long purchase cycles limit the conclusions.
The ten metrics at a glance
| Metric | Use and limitation |
|---|---|
| MER | Total revenue / all marketing costs; does not show profit. |
| New-customer CPA | Defined media spend / new buyers; calculate complete CAC separately. |
| EMQ | Matching diagnosis; not proof of revenue. |
| Blended ROAS | Total revenue / media; not proof of incrementality. |
| Hook rate | 3-second plays / impressions; compare like conditions. |
| Hold rate | ThruPlays / 3-second plays; record length and definition. |
| Outbound CTR | Outbound clicks / impressions; not a page arrival or sale. |
| Placement CPM | Price per 1,000 impressions; add purchases and contribution. |
| Frequency | Impressions / reach; state the period. |
| New/returning ROAS | Cohort return with justified allocation; not automatic advertising lift. |
Hook, hold and CTR are percentages; EMQ is a diagnostic score, not a percentage.
One screen for the business owner
Start with confirmed net revenue, contribution, returns, new buyers, CAC, MER, media spend and clearly labelled platform attribution. Show CPA from Meta alongside CPA from actual orders. Growing differences can reflect events, currencies, cancellations or attribution.
Below, show the diagnosis: what changed and which test is justified next? Technically, split the funnel into product views, cart additions, checkout and confirmed purchase. Check transition rates separately by mobile device, country and language when those differences change decisions. Avoid changing campaigns, creatives and the website simultaneously if you need to identify the effect.
Frequently Asked Questions
Why use MER alongside ROAS?
MER connects confirmed overall revenue with all defined marketing costs. Platform ROAS uses attributed revenue and that platform’s media spend. Both are useful, but neither MER nor ROAS replaces a contribution calculation.
What Event Match Quality is good enough?
There is no universal score that guarantees profitability. For each event, check permitted parameters, delivery errors, deduplication and the actual order. EMQ supplements that diagnosis rather than replacing it.
What is the difference between hook rate and hold rate?
Hook rate divides 3-second plays by impressions. Hold rate divides the selected ThruPlays by 3-second plays. Both ratios require comparable video lengths and placements, and neither establishes a sale.
Why separate ROAS for new and returning buyers?
First and repeat buyers have different costs and commercial significance. Purchase history must come from the shop or CRM, with a justified allocation of revenue and spend. The split alone does not establish additional sales created by advertising.
What is a good ROAS?
A value that delivers the required contribution after product costs, returns, delivery and other expenses. The same ROAS can work for one category and be inadequate for another. There is no universally valid value.
Can advertising be managed using Meta data alone?
It is useful for daily campaign work. Profit and budget decisions also need shop, CRM and financial data with comparable definitions. The platform does not establish complete company profit.
Should countries be assessed separately?
Yes, when currency, prices, delivery, margin or the buying journey changes the decision. Combine markets only when their differences do not conceal a relevant decision and individual segments provide too little data.
Meta reports growth that the business cannot see? Discuss the figures that disagree or read about our Meta Ads service. Start with both data sources and a concrete question; further diagnostic work is agreed separately.
