How to calculate ROAS, CPL and LTV/CAC in 2026: formulas, errors, niche calculator

In 80% of client reports that I review on discovery calls, ROAS is calculated incorrectly. Not “approximately”, not “with an error” - literally according to a formula that gives a figure 30-60% higher than the real one. This is not a theoretical article about metrics. These are six formulas, six common mistakes and a table of reference values ​​for 5 niches, on which you can check your calculation right now.

The metrics in this article are for a performance marketer who runs campaigns himself, and not for an analyst. There are no regression models or advanced econometrics here. Only something without which it is impossible to honestly say “channel N is profitable” or “the budget needs to be reallocated.”

A performance metric without a correctly calculated denominator is not a metric. This is self-deception in a beautiful wrapper.

1. CPL - cost per lead

The most used and most distorted metric. Formula:

CPL = Cost per channel / Number of leads from this channel

What is a “lead”? Determine on the shore. Options:

• Completed form (low barrier, high quantity, low quality)
• Call or chat message (medium barrier)
• Request a demo or consultation (high barrier)
• MQL (marketing qualified lead) - a lead that has passed the initial qualification

The most common mistake: counting CPL based on “general leads” from CRM, without dividing it by source with UTM. It turns out like this: a channel with 80% of garbage leads and a channel with 20% of quality ones - both show “CPL 300 ₽”. Solution: calculate CPL only by MQL, and not by the total number of applications. Details in my article about attribution 2026.

Niche guidelines are in the table below.

2. CPC - cost per click

CPC = Cost per channel / Number of clicks

A metric that in itself doesn’t say anything. CPC 80 ₽ - is it good or bad? Depends on CTR, click quality, and LP conversion. CPC is only useful in conjunction with CPL and CR (conversion rate).

The main trap is optimizing only to reduce CPC. A cheap click does not guarantee a cheap qualified lead. Compare campaigns by CPL and subsequent sales, and use CPC as a diagnostic metric for the auction.

3. CPA - cost of target action

CPA = Cost per channel / Number of targeted actions completed

CPA differs from CPL in that the “action” is not a lead, but a purchase/subscription/any conversion with economic value. CPA is the final indicator of the funnel, not the middle one.

The connection between CPL and CPA through the conversion rate of the funnel:

CPA = CPL / Lead-to-customer conversion rate

Example: CPL = 300 ₽, CR from lead to client = 15%
CPA = 300 / 0.15 = 2,000 ₽

4. ROAS - return on advertising costs

ROAS = Revenue per channel / Expenses per channel

ROAS shows how many rubles of revenue each ruble of advertising budget brings. ROAS 3.0 = 3 ₽ revenue per 1 ₽ advertising.

The main errors in calculating ROAS:

• They count based on the last click, without taking into account multi-touch. The ROAS of the first-touch channel may be 0.5 (because another channel gives a direct conversion), but without it the entire funnel will fail. The solution is MMM-light or incrementality tests.
• They take “turnover”, not “revenue”. If you have a refund rate of 12%, ROAS should be calculated from net revenue, not from payments.
• The long cycle is not taken into account. ROAS on a monthly horizon distorts the picture for products with long-tail purchases. Calculate ROAS for 30/60/90 days.

The difference between ROAS and ROI:

ROI = (Revenue - Expenses) / Expenses * 100%
ROAS = Revenue / Expenses

ROI takes into account profit, ROAS only takes into account revenue.
ROAS 3.0 is not equal to ROI 200% - you need to subtract the cost of the product.

5. LTV - customer lifetime value

LTV = AOV × Frequency × Customer lifespan

where:
AOV (average order value) - average bill
Frequency - average number of purchases per period (year)
Customer lifespan - average duration of relationship with a client, years

In reality, LTV is calculated in two ways: historical (based on past revenue) and predictive (modeling the retention curve). For marketing it is quite historical - you take a cohort of customers who first bought N months ago, count how much they have brought in since then.

The main mistake: using “average LTV for all clients” without segmentation. The LTV distribution is always heavily skewed to the right: 10% of customers account for 50–70% of revenue. Segmenting by behavioral cohorts is the only way to deal with LTV fairly.

6.CAC and LTV/CAC ratio

CAC (customer acquisition cost) = All acquisition costs / Number of attracted customers

LTV/CAC ratio = LTV/CAC

Healthy LTV/CAC = 3+ (for 1 ₽ of attraction - at least 3 ₽ of value for the entire time)
LTV/CAC < 1 - the business loses money on each client
LTV/CAC > 5 – you are underinvesting in marketing

Critical: CAC is not equal to CPA. CPA is the cost of one transaction. CAC is the cost of attracting a customer who may make many transactions. The difference is in the denominator. In recurring businesses (SaaS, subscriptions), the difference can be 5–10x in favor of CAC.

Full LTV/CAC benchmarks by niche - in my separate article about unit economics in Russia.

Payback period - how many months for CAC to be returned?

Payback period = CAC / (Monthly revenue per customer × Gross margin)

Healthy payback < 12 months for SaaS
<6 months for e-com with an average bill of up to 5K ₽
< 3 months for impulse purchases

Payback is what separates “marketing works in the long run” from “marketing eats up cash flow.” If you have a payback of 18 months on an e-com product, the business has a serious problem with unit-economics, which cannot be solved by “just increasing the conversion.”

Reference values for 5 niches in the Russian Federation

The numbers are medians from my 9 projects 2024–2026 + an anonymous pool of 12 colleagues.

If your numbers are far outside the range, this is not a reason to panic, but a reason to check two things: 1) the correctness of attribution, 2) the correctness of segmentation. On case with segmentation failure −20% I had exactly this situation: the middle ones looked “like everyone else,” but inside there were two groups with radically different economies.

6 calculation errors that I see in 80% of client reports

1. CPL for “all applications”. Without separating junk-leads from MQL, the figure lies by 30–80%.
2. ROAS is based on “turnover” rather than net revenue. Refunds, refusals, missed calls - must be deducted.
3. LTV is “average across all clients.” The distribution is skewed and the mean is not representative. Calculate the median by segment.
4. CAC = advertising costs only. The salaries of the marketing team, tools, and content production should also be included. Otherwise, CAC is underestimated by 1.5–2×.
5. Payback on the “average check”. It is necessary for marginal revenue (taking into account margin).
6. ROAS on the monthly horizon. For long-cycle products - count for 60/90/180 days.

If you need automated calculations

I keep everything in one Google Sheet with formulas and Ya.Metrika API-feed. Uploaded weekly via n8n (details in my article about n8n agents). I can send you a table template upon request - I usually show it at the discovery call and adapt it to your case.

If you're not confident in your numbers right now, that's okay. Most companies think “somehow.” Would you like me to analyze your current measurement for 30 minutes? discovery — I’ll show you where the formulas make mistakes and what 3 corrections to make in the first week.

Related materials: ROAS/ROMI/DRR calculator, CPL calculator, CPA calculator, ROAS benchmarks for e-com.