One CRO proposal offers an audit. Another includes design and development. A third sells a monthly testing programme. Comparing the final amounts will not tell you which one resolves the problem on your website, because they may end at different points.
There is no defensible single price for “CRO” without the work behind it. Ask for a quote that separates investigation, the proposed change, implementation and evaluation. Our recommendation is to buy the smallest complete work package that lets the business make a useful decision, then expand if the evidence supports it.
Define what the customer should be able to do
Start with a specific route: choose a product variant, calculate delivery, complete payment or submit a suitable enquiry. “Improve the website conversion rate” can cover several different problems, each requiring different work and access.
Record the starting event, the intended outcome and the visitors included in the comparison. A page may attract buyers, existing customers and job applicants. A change in their proportions can move the headline conversion rate without any change in usability.
For B2B, specify whether the desired outcome is a submitted form, an accepted enquiry or an attended meeting. For a shop, decide how cancellations and returns enter the assessment. These choices affect both the quote and the evidence needed to accept the work.
Separate the costs that proposals often combine
- Investigation: measurement checks, funnel review and examination of relevant customer behaviour.
- Design: a proposed change, supporting copy and the agreed screen states.
- Implementation: development, integration, accessibility checks and release preparation.
- Evaluation: a comparison plan, analysis and the decision about further work.
- Additional costs: tools, participant recruitment, content production and support where required.
Some businesses need all of these. Others already have a developer and need a well-supported diagnosis plus an implementation brief. The cheaper proposal may leave development with your team; that can be appropriate if the responsibility and cost are visible.
Ask what is delivered if investigation finds no evidence for the proposed redesign. A useful supplier can recommend keeping a working section, repairing measurement or examining the offer instead. An obligatory redesign makes the investigation less useful as a purchasing decision.
Check whether the investigation can support a conclusion
An audit should name the pages, devices and customer routes it covers. It should distinguish an observed problem, a hypothesis about the cause and an untested suggestion. A list of preferences about colours or button text is not enough to justify development.
Ask how observations will be selected. A recording of one confused visitor can expose a fault worth investigating, but it does not establish how common the fault is. A heatmap also needs context: the page version, visitor segment and action being examined.
When event collection is inconsistent, include the cost of checking it. The GA4 event documentation explains event collection; the presence of an event alone does not establish that it represents a delivered enquiry or a retained order.
Use arithmetic to compare scope, not invent a market average
Consider a purely illustrative quote: 16 hours of investigation at €100 per hour cost €1,600. Another 8 hours at €100 cost €800, making €2,400 in total. These assumed amounts are neither market benchmarks nor Canem Errant prices.
The useful question is what the second block buys. Does it include development, review and corrections, or only a design file? Are subscriptions, VAT, additional languages and deployment excluded? The same total can describe materially different work.
For a separate illustration, assume the complete project costs €2,400 and each additional retained order contributes €30 after relevant variable costs, with advertising costs otherwise unchanged. Recovering the project cost requires 80 additional retained orders under those assumptions. This is a break-even calculation, not a prediction that CRO will produce those orders.
For lead generation, the equivalent calculation needs lead acceptance, sales conversion, contribution per sale and the time to close. Treat an unknown input as unknown. Filling it with an attractive industry percentage makes the quote look more certain than the business case.
Decide whether a test is feasible before buying one
A monthly A/B testing programme needs sufficient eligible observations for the proposed questions. Ask for the baseline, the effect the test is designed to detect, the expected sample and a stopping rule. A promise to “test every week” says little about whether the results will support decisions.
With limited traffic, a narrower approach may be more useful: repair a reproducible fault, observe representative tasks and document what still requires measurement. This is different from declaring a small before-and-after movement a proven conversion lift.
Test integrity also matters. Microsoft Research describes sample ratio mismatch as a diagnostic problem in experiment allocation. A polished results chart should not hide a broken comparison. Our A/B testing guide develops the planning questions further.
Agree acceptance before development starts
Separate technical delivery from business effect. A corrected form can be checked against agreed behaviour and successful delivery; attributing an increase in qualified leads requires a suitable comparison and enough observations.
Define supported screen sizes, error states, event behaviour, review responsibility and the process for correcting defects. If several languages are in scope, include their content and interface checks explicitly. “Responsive design included” is not a precise acceptance criterion on its own.
Assign deployment authority and record a recovery plan for the changed route. Include dependencies on the shop platform or the company’s developers. A supplier can deliver its agreed work while a third-party release remains pending; the status should make that distinction clear.
Choose a project or continuing work for a reason
A bounded project suits a known obstacle with a manageable route to investigation and correction. Continuing work can suit a business with recurring hypotheses, suitable traffic and someone responsible for implementation. Neither payment model is inherently more effective.
Ask how priorities will be revised when the first hypothesis is rejected. A continuing contract should not turn an empty test queue into unnecessary website changes. A project should not end with a presentation when the buyer expected implementation.
Canem Errant’s CRO service starts from the obstacle between a click and a useful outcome. Bring the page, the expected action and the available data. The investigation, agreed changes and evaluation can then be priced with clear responsibilities.
Frequently Asked Questions
Can you quote CRO without access to analytics?
You can scope a public review or a limited investigation, with its limits stated. A precise diagnosis or reliable prediction of improvement needs more evidence than a website address alone.
Is an audit enough to improve conversion?
An audit can establish what to investigate or change. The effect depends on implementation, the offer, traffic and measurement. Ask who will deliver the change and how it will be assessed.
Should a CRO agency guarantee a percentage improvement?
A supplier can commit to defined work and acceptance criteria. A universal conversion lift cannot be established before understanding the business and the evidence. Ask for a bounded test and an honest stopping condition.
