Conversion rate optimization, or CRO, is the practice of improving the experience so more eligible visitors complete a valuable action. In ecommerce, that action is often a purchase, but intermediate actions can help explain the journey.
The work begins with understanding an obstacle. A new button color is a possible change; “customers cannot tell whether the kit includes the tools they need” is a problem worth investigating.
Define the outcome and its denominator
Choose a primary outcome and describe how you count it. A session purchase rate measures purchasing sessions divided by eligible sessions. A user purchase rate uses people as the denominator. An add-to-cart rate measures a different step.
Do not switch definitions halfway through a comparison. Use intermediate metrics to diagnose behavior, then check the business result. More add-to-cart events are not automatically valuable if completed purchases decline or returns increase.
Track a few measures that would reveal harm: contribution per visitor, refund rate, support contacts, or checkout errors, depending on the change. The point is to improve a worthwhile outcome without hiding the cost elsewhere.
Combine the numbers with direct observation
Start with reliable analytics. Identify where behavior differs by device, product, or traffic source. Then look for a plausible explanation through support questions, customer interviews, usability sessions, or appropriately configured session recordings.
Ask a participant to complete a realistic task: “Choose a kit you could give a friend who has never tried embroidery.” Observe where they hesitate and which information they seek. Avoid telling them what you want them to notice.
A heat map can show where recorded interactions occurred; it cannot reveal a shopper’s motivation by itself. A support question can suggest uncertainty; it does not establish how common that uncertainty is. Use the methods together and record what each can actually support.
Write a hypothesis that connects problem and change
For a hypothetical store, the evidence might be repeated questions about what comes in a kit and observed difficulty finding the contents list. A useful hypothesis is:
If we put a clear included-versus-needed list near the purchase controls, first-time visitors will understand the offer more easily, which may increase completed purchases without increasing product-related returns.
This tells the team what to change, why it might work, and what outcome to watch. “Make the page more modern” does not.
Choose one coherent change when you want to learn about a particular explanation. If you redesign the entire page, you may learn whether the new package works, but you will have less evidence about which component produced the result.
Choose an appropriate evaluation method
An A/B test assigns eligible participants to different versions and compares outcomes. The UK government’s account of using A/B testing explains why random allocation helps distinguish a design effect from other changes over time.
Before launching a test, define the hypothesis, assignment unit, main metric, minimum effect worth detecting, planned sample or stopping rule, and monitoring for serious failures. Use a suitable experiment calculator or statistical method with your actual baseline. There is no universal number of visits that makes every test reliable.
Keep assignment consistent so people do not repeatedly switch versions. Check that the experiment itself works, including allocation and measurement. Avoid declaring a winner simply because a dashboard briefly crosses a significance threshold during repeated checking; use the test method’s planned decision rules.
What if your store has little traffic?
Fix obvious defects such as broken controls, incorrect facts, or unreadable text. You do not need an experiment to justify making a payment button function.
For uncertain design questions, use focused usability sessions and customer conversations. These can reveal problems and improve your understanding, but a small qualitative sample cannot estimate a population conversion lift.
If you release a change and compare before and after, document other factors: discounts, channel mix, stock, and seasonality. Treat the result as directional evidence unless the evaluation design supports a stronger conclusion.
Interpret the result in business terms
Suppose a fictional test reports a move from 2% to 2.2%. That is a 0.2 percentage-point increase and a 10% relative increase. Whether the evidence is persuasive depends on sample size, uncertainty, test design, and data quality.
Even a credible purchase lift needs an economic review. If it comes from a discount, check whether the additional contribution covers the reduced margin. A test can succeed on conversion rate and disappoint on the business outcome.
Use AI to sharpen the experiment brief
Here are our observed issues, metric definitions, and traffic constraints: [inputs]. Propose three testable hypotheses. For each, separate evidence from assumptions, define one coherent change, name a primary outcome and possible harms, and recommend whether usability research or an experiment is the better next step. Do not invent effect sizes or claim statistical significance.
Your assignment: write one experiment brief with the evidence, proposed change, measurement plan, and decision rule. If research is the right next move, write the participant task instead. Use the website maintenance guide to prepare and verify the change before exposing shoppers to it.