Ece runs e-commerce for a home textiles brand, and three CRO proposals sit on her desk. The first promises four tests a month for a fixed monthly fee. The second asks for a small fixed fee plus a share of the conversion lift. The third proposes a one-off audit and gives no figure for what comes after. The three proposals are priced in three different units; Ece cannot even tell which one is expensive.
I invented Ece for this article; I did not invent her question — "what does CRO consultancy cost?" comes up in the first ten minutes of every buying conversation. The honest answer is not a number but the five variables that produce one: traffic volume, the state of your measurement, scope, who holds the development resource, and reporting. Below I take the five one by one, say which items on our own price list correspond to CRO and which do not, and then show how to work out the return on the price with your own traffic and order value.
This article is the "what does it cost" part of the CRO decision set. The criteria for shortlisting an agency are in how to choose a CRO agency; how the work itself runs — how we work as a CRO agency — is on the service page.
What sets the price of CRO consultancy?
Five variables set the price of CRO consultancy: the traffic available for testing, how much the measurement needs repairing, the scope of the work, who builds the winning changes, and how deep the reporting goes. Two proposals from the same agency to two stores can differ several times over because one of the five differs — and most of the gap between agencies comes from the same place.
Internationally, there is one compilation that shows how wide the gap runs. "How Much Should a CRO Agency Charge?", published by the A/B testing tool maker Convert on 6 August 2025 and updated on 3 April 2026, summarises CRO agency fees under a monthly retainer as $2,000 at the low end, $16,000 on average and $30,000 at the premium end, and a one-off conversion audit at $2,500 to $4,500. The figures come not from a survey but from tier ranges published by Invesp, itself a CRO agency, and from expert opinion, and they are in dollars. They are useful not for benchmarking proposals in Türkiye but for seeing why the same job title can be priced fifteen times apart. We could not find a CRO price compilation for Türkiye with a published method.
The five sections below trace where that gap comes from, one by one. At the end of each I have written the line to look for in your proposal.
Why does traffic volume change the price?
Traffic decides how many tests can reach a meaningful result in a month, and what a monthly fee buys is learning per test. On a site with few visitors, the same fee buys fewer tests, and so more expensive learning.
Put numbers on it. On a page converting at 2%, telling a 10% relative lift — from 2% to 2.2% — apart from noise at 95% confidence and 80% statistical power takes roughly 80,700 visitors per variant, around 161,000 across two variants. If the page gets 100,000 visitors a month, the test runs about seven weeks. At 20,000 a month it runs eight months — in practice it never finishes, because in eight months the season, the prices and the campaign calendar all change and the test starts measuring something else.
On small traffic, the way out is to look for bigger effects. Looking for a 20% relative lift on the same page cuts the visitors needed to roughly 42,000; at 20,000 visitors a month that is two months. But big effects only come from big changes: a rebuilt checkout flow, a different way of presenting price and delivery charges, a product page reorganised from the top. A button-colour test is the most expensive thing you can buy on small traffic.
Traffic moves the price in two directions. With little traffic the programme shrinks but the cost per test grows; so on a low-traffic site the right purchase is not a testing programme but a direct audit, with obvious blockers fixed without waiting for data. With plenty of traffic several tests can run at once, and the price grows with analyst and development hours instead. The line to look for in the proposal: the expected duration per test, and the traffic assumption behind it.
What does broken measurement add to the budget?
Broken measurement adds a line to the budget that comes before any test: redefining the conversion events, removing double counting, passing order value to analytics correctly, and verifying mobile and desktop flows separately. Skip it and the rest of the budget is spent on faulty data.
The cost of this line mostly shows up in the calendar. Typically the first two weeks go to validating measurement; if the setup is badly broken this stretches and the first test slips with it. A cheap-looking proposal that says "we'll use your current setup" may have become cheap by deleting this line — and the deleted line leaves every test result in doubt.
That was the order in the GYMWOLVES case: the data flow was repaired first, the conversion funnel rebuilt next, and the campaign fed last with social proof shot with athletes; by the end of the third month sales were up 12×. What made that result readable was measurement being the first line. If you want to see your own position before asking for proposals, Diagnoo does the tag part of this check at no cost: it looks at whether GA4, Meta Pixel and session-analytics tags are installed on your store. The line to look for in the proposal: the scope and duration of the measurement repair, and whether it is inside the price.
How do audit, testing programme and implementation split the price?
CRO scope has three layers, and each is priced differently. The audit is one-off and suits a fixed price. The testing programme is continuous and priced monthly. Implementation — building the winning changes permanently — is missing from most proposals altogether and produces the biggest surprise.
An audit delivers a map of the losses and a prioritised test backlog: where visitors give up, the estimated loss at each point, and which hypothesis to try first. I have written up how it is done, step by step, in the e-commerce GAP analysis guide. If you want to see where your own rate sits against your sector before an audit, the article where we gathered e-commerce conversion rate benchmarks is the right starting point for the comparison.
Let me use our own price list as the example. Our product closest to the audit layer is the Growth Sprint: it runs four weeks at a fixed €7,500 excluding VAT. Its scope includes a conversion rate benchmark per channel, an A/B test plan with at least three hypotheses for the single most critical funnel step, and an eight-week test backlog in which each test is written with its hypothesis, priority score and expected conversion impact. To be plain about it: the Growth Sprint is not a pure CRO package — it combines a channel audit with a funnel diagnosis and does not cover a continuous testing programme. Quick wins such as a landing page test can go live within the sprint; running the backlog falls, after the sprint, to your team or to a separate project or monthly engagement. Media budget, tool licences and creative production also sit outside the price.
We have no published monthly price for a continuous testing programme. A monthly fee written before the traffic and development questions are answered means the scope later either shrinks or grows; either way, the invoice rather than the work ends up at the centre of the relationship. The line to look for in the proposal: which of the three layers are included in the price, by name — implementation above all.
Who should hold the development resource?
The development resource is the most hidden line in a CRO budget. Setting up a test is one job; making the winning variant a permanent part of the site is another — and if the second is not in the proposal, it is paid for either in your team's hours or in the gain arriving late.
There are three models. In the first, the agency builds both the test variants and the permanent change; the fee is higher but the gain goes live quickly. In the second, the agency handles hypotheses and analysis and your team takes on development; the fee drops, but your team's backlog sets the pace of testing. The third is a hybrid: the agency builds variants inside the testing tool and your team makes the winner permanent. For most mid-sized stores the third is the balanced option — provided the time to ship a winner is written into the contract.
The platform enters the calculation too. On hosted platforms such as İKAS, Ticimax, İdeaSoft or Shopify, changes to the theme and the checkout step go only as far as the platform allows; some hypotheses cannot be tested at all, and some only through an add-on. With your own software anything is possible, but every change is development time. If a winning variant waits three months in the queue, three months of lift never happened; the return calculation has to include that delay. The line to look for in the proposal: the time to ship a winner, and whose budget that work comes out of.
Why is reporting a separate line in the budget?
Reporting is analyst time, and analyst time takes up budget. Each test's hypothesis, duration, sample collected, outcome and the decision it changed go into a written record; without that record the programme tests the same idea a second time next quarter.
A cheap proposal gets cheaper by reducing reporting to a dashboard. A dashboard shows what happened, not why; and losing or inconclusive tests never appear on it. Ask for the format and frequency of reporting at proposal stage: a monthly results report, a one-page record at the end of each test, or a roadmap review every three months? Each carries a different analyst load, and so a different price.
Ownership is part of reporting too. The testing tool account, the conversion definitions and the archive of past tests should sit in your name; paying extra to get them back when the engagement ends is the most expensive form of reporting there is. The line to look for in the proposal: the report format, its frequency, and who keeps the test records when the engagement ends.
Should you pay per project or per month?
Both are right, depending on the phase of the work: the first audit and the measurement repair run as a project, the continuous testing programme runs monthly. Reversing the order — starting a monthly programme without an audit — means paying the first three months' fees without knowing which page is being tested, or why.
Look for three conditions before moving from a project to a monthly engagement. Traffic should be enough to finish at least one test a month in a reasonable time. There should be development capacity to ship a winning variant within two weeks. Measurement should be validated and the baseline period on record. Without all three, a monthly fee is a fee for tests that are waiting.
Make what the monthly fee buys countable in the contract: how many tests start each month, what the average test duration is expected to be, how many hours of development are included, and how often the roadmap is re-prioritised. The method behind our conversion rate optimisation service is built on the same logic: first the leaks are found, then the hypotheses are prioritised, then they are tested and measured, and last the routine is handed over to your team.
Why is performance-based pay risky?
Performance-based pay means paying — or not paying — for variables the agency does not control. The conversion rate also moves with the season, the campaign calendar, price changes, stock levels and the traffic mix; if the contract cannot separate the part of that movement that belongs to the agency, the fee has been tied to noise.
Three problems recur. The first is the baseline fight: against which period is the lift measured, and does the sale week count as part of the base? The second is incentive drift: a fee tied to conversion rate rewards moves that lift the rate by paying for it out of margin — aggressive discount codes, a lower free-delivery threshold, tests kept short and declared winners early. The third is the price itself: an agency that takes on the risk prices that risk in, and in a good year the total fee comes out higher than a fixed one.
If you still want the model, put four conditions in writing. The metric should be revenue per visitor or contribution per order rather than conversion rate, because the rate can rise at the expense of margin. Measurement should use a control group in which a small share of traffic stays on the old version. The baseline period and the measurement source should be fixed before signing. The variable fee should have a cap and sit on top of a fixed base; a contract tied entirely to success pushes the agency towards tests that are easy and quick to win.
What should you watch for in a cheap CRO proposal?
A cheap proposal usually gets cheap by deleting a line; the question is which line. Look in five places.
- A test count but no duration: if "four tests a month" is written without a traffic assumption and a sample size calculation, most of those tests will measure noise.
- No measurement repair: "we'll start with your current setup" can mean the line that should come before the first test has been deleted.
- No development: only the text and colour changes the testing tool's visual editor allows get tested, and the structural changes that produce big effects stay out of scope.
- An unclear licence: if the testing tool licence is not in the proposal it will be invoiced separately; if the account is in the agency's name, the past records leave with the agency when the engagement ends.
- A report that carries only winners: a report in which losing tests never appear spends next quarter's budget on the same ideas.
One more warning: advertising management proposals sometimes carry a line reading "conversion-focused improvements". That line is an improvement item that comes with the advertising, not a CRO programme. If there is no sample size calculation, hypothesis backlog or testing calendar behind it, counting it as your CRO budget means paying a CRO price for work nobody measures.
Work out the real price of a cheap proposal in one line: the proposal amount, plus your team's development hours, plus the separately invoiced tool licence, plus the cost of a test wrongly declared a winner settling into the live site. The last item looks like zero, because nobody ever invoices it.
How do you work out the return on a CRO price?
Rather than estimating the lift you expect, calculate the lift the price needs in order to pay for itself; the second needs only your own numbers and involves no forecasting. It takes four steps: monthly orders, contribution per order, the monthly equivalent of the proposal, and the number of extra orders that would cover it.
The figures below are hypothetical; they belong to none of our clients and were chosen only to show how the calculation is built. The store gets 60,000 sessions a month and converts at 1.5%: 900 orders a month. The average order is €40; after product cost, delivery and payment fees, the contribution per order is 35%, or €14. For the cost side, let us use a real figure from our own list: €7,500, the list price of the Growth Sprint. Assuming the gain lasts twelve months, that is €625 a month.
Divide €625 by €14: the price needs roughly 45 extra orders a month to pay for itself. Against 900 orders that is a 5% relative lift; the conversion rate has to move from 1.5% to about 1.58%. On this traffic the difference is too small to prove with a single test — it would need more than 400,000 visitors per variant — but as the sum of several fixes it is a reasonable target. That is my reading.
Apply the same calculation to a store with 15,000 sessions a month. Orders drop to 225, and the same 45 extra orders now mean a 20% relative lift: the rate has to move from 1.5% to 1.8%. Telling a 20% difference apart in a single test takes roughly 28,300 visitors per variant, around 56,600 in total — close to four months on this traffic. At the smaller store the same price demands a relative lift four times as large; that is the traffic section's point, in numbers.
Run the calculation with your own numbers and add two things. The first is delay: the later a winning change goes live, the shorter the twelve-month assumption becomes. The second is margin: a change that buys the lift with a discount raises the rate but lowers the contribution, so the unit of the calculation should be contribution per visitor rather than conversion rate. Which tactics can produce that lift is covered in the article where we set out twenty-one tactics for raising your conversion rate.
How do budget and scope change with the size of the store?
Scale changes the weight of the five variables. In a small store the binding constraint is traffic; in an exporting store it is the number of markets; in a large retailer it is development and internal sign-off. The three profiles below give no figures; they describe where the budget should go.
An SME store
For an SME store at around 15,000 sessions a month, buying a testing programme is usually premature; the calculation in the previous section shows why. The budget belongs in a one-off audit and in fixing the obvious blockers — forced registration, a long form, a delivery charge that appears late, a slow mobile page — without waiting for data. We do not recommend the Growth Sprint to brands with annual revenue below 20M TL, because at that scale the gap a channel audit can open stays narrow. At this size a free diagnosis and the basic fixes your own team can make earn more than a paid programme.
An exporting store
In an exporting store, traffic splits across markets and each market is tested on its own sample; traffic that looks high in total can be small per market. Measurement has to be validated per market as well: currency, payment methods and the way delivery and customs charges are shown are separate leak points in each one. Hypotheses differ by market too — in one market a missing trust signal loses the sale, in another the delivery time does. The budget grows with the number of markets, but not in a straight line: the right order is to build the method in the largest market and then validate the winning patterns in the others.
A large retailer
In a large retailer traffic stops being the constraint, and several tests can run at once. The budget now grows with testing velocity: how many tests a month can be built, pass quality assurance and go live. An enterprise testing platform licence is a separate line, and development, quality assurance and internal sign-off can take up more room than the agency fee. At this scale the value of what you buy from outside lies not in the number of tests but in programme design, hypothesis quality and the routine handed over to the in-house team.
Conclusion: which calculation should you run before approving the budget?
The price of CRO consultancy is not a label but the result of five variables: traffic, measurement, scope, development resource and reporting. Before comparing two proposals on price, put those five lines side by side; most of the gap shows up there, and it usually points to the line the cheaper proposal deleted.
Here is the concrete test you can run today: note your average monthly sessions over the last three months, your conversion rate, your average order value and your contribution rate per order. Divide the proposal on the table by twelve, then by the contribution per order, and set the resulting number of extra orders against your current monthly orders. Roughly speaking, if the percentage comes out around 5% you are discussing a reasonable target; if it is above 20% you first need traffic or a bigger change, not a testing programme.
Run the same calculation on our proposal too. The scope of our CRO consultancy service — what is included and what is not — is written out on the service page; put the numbers side by side and compare.