An ecommerce conversion rate optimisation agency sells improvement to the part of your funnel you already paid for: the visitors who arrive and leave without buying. Done properly it is the cheapest growth available, because nothing about acquisition has to change. Done badly it is a series of button colour tests with confident conclusions drawn from noise. The difference is almost entirely about whether the agency has the discipline to admit when a test proved nothing. This guide covers when hiring one makes sense, what volume you need first, how fees are structured, and how to vet a candidate on evidence.
You need enough traffic before testing is possible
Split testing is a statistical exercise, and statistics need volume. A store with a small number of orders per month cannot run a test that reaches a trustworthy conclusion in a reasonable time, because the difference you are trying to detect is smaller than the natural week-to-week variation in your own sales. Below that threshold the honest work is qualitative: session recordings, on-site surveys, usability testing, checkout instrumentation and fixing the obvious breakages, none of which needs a control group. A good agency will tell you which of the two regimes you are in during the first conversation, and will decline the testing programme if your volume cannot support it. An agency that proposes a twelve-month testing roadmap without asking your order volume is selling a process rather than an outcome.
What the work actually consists of
A competent engagement runs a loop. It starts with instrumentation, because most stores cannot yet see where the funnel leaks: checkout step tracking, error logging, device and traffic source splits. Then research, combining analytics with recordings, surveys and support tickets to generate hypotheses about why people leave. Then prioritisation, which is where judgment shows: the best agencies test the things that would matter if true, not the things that are easy to build. Then implementation and measurement, with a predefined sample size and duration agreed before the test starts, not after the results look interesting. Then a decision, including the decision that a test was inconclusive. Ask any candidate to describe a test that failed and what they did next. The answer separates practitioners from vendors.
Fee models, and the trap in performance pricing
Three structures are common. A monthly retainer buys a named team and an agreed test cadence, and is the most straightforward to compare. Project pricing suits a one-off audit and a fixed set of fixes, and is a sensible way to trial an agency before committing. Performance pricing, a share of incremental revenue, sounds perfectly aligned and rarely is, because incremental revenue has to be defined, attributed and agreed, and seasonality, promotions and traffic mix all move the baseline. If you go that route, agree the measurement method in writing before the first test, including who calculates it. Whichever model you choose, ask what is excluded: design, development, and the cost of the testing platform itself are the usual extras, and ecommerce operators comparing agency retainers should price those in before comparing headline numbers.
How to vet a candidate
Ask for a real test report, redacted if necessary, and read how the conclusion is worded. You are looking for a stated hypothesis, a predetermined sample size, a confidence figure and a plain statement of what the result does and does not prove. Ask what share of their tests are inconclusive; anyone claiming a high win rate is either lucky, working on badly broken sites, or calling noise a victory. Ask who implements: an agency that cannot ship code will be blocked by your developers all year. Confirm you own the testing account, the research artefacts and the code. And check the basics that apply to any supplier: published pricing or a disclosed minimum, contactable clients, and a named team that stays after the pitch.
Questions people ask about ecommerce conversion rate optimisation agency
How much traffic do I need for testing to work?
Enough that a meaningful improvement would show above your normal weekly variation within a few weeks. The practical test is arithmetic on your own order volume, and any competent agency will run it with you before quoting. If the answer is no, spend on qualitative research and fixing known breakages instead.
Will CRO hurt my search rankings?
Not if it is implemented properly. Google's guidance on testing is that ordinary A/B testing is acceptable when the same content is served to users and crawlers, tests run only as long as needed, and redirects used for testing are temporary. Cloaking different content to search engines is the line you must not cross.
Retainer or project?
Start with a project: an audit plus a defined set of fixes tells you how the agency thinks, at a bounded cost. Move to a retainer only if your volume supports continuous testing and the audit showed real judgment. Long retainers on low-traffic stores are the most common waste in this category.
What is a realistic result?
Compounding small gains rather than one dramatic jump. Most durable improvement comes from removing friction in checkout, clarifying delivery and returns information, and fixing mobile problems. Be sceptical of any agency quoting a headline uplift figure from another client's store as if it predicts yours.