OMCA: Mobile Checkout Friction and an Overloaded Redesign

Mobile Checkout Friction and an Overloaded Redesign

Lumen Home sells lighting online. Mobile accounts for 78% of sessions but only 61% of revenue. The checkout has four steps, requires account creation unless users find a hidden guest link, and collects 14 fields, including optional phone and company name. Cart abandonment is 62%, and 22% of mobile checkout sessions hit an address-validation error that clears the form. Largest Contentful Paint is 3.1 seconds on mobile, and the hero image is 1.8 MB. Analytics shows a 34% drop-off between shipping and payment. The team has 40,000 monthly sessions, a 1.8% baseline conversion rate, and can run one A/B test at a time with 95% confidence and 80% power. The product manager wants to redesign the entire checkout in six weeks. The UX researcher has five days and can recruit eight users. What should the team do?

1. Prioritize the checkout problems. Which issue should the team test first and why?

Read the worked answer

Prioritization

Use a transparent framework such as impact, confidence, and effort. The biggest observed friction is not the number of fields alone; it is the address-validation error that clears the form and the hidden guest option. Both create hard stops.

Address-validation failure: high impact because it affects 22% of mobile checkout sessions, high confidence because the error is observable, and medium effort to fix. This is likely the first action because it removes a known failure, not a hypothesis.
Guest checkout: high impact, high confidence, low effort. Expose guest checkout as the default or equal option. Required account creation is a known abandonment driver, especially on mobile.
14 fields: reduce to essential fields. Optional phone and company name should be removed or moved post-purchase. Medium impact, high confidence, low effort.
Four steps: collapsing shipping and payment may help, but a full redesign is high effort and confounds variables. Test step reduction after fixing hard errors.
LCP 3.1 seconds and 1.8 MB hero: performance matters, but checkout speed is more important than the home hero. Compress checkout assets and measure checkout LCP separately.

First action

Fix the validation error for everyone and measure before and after rather than A/B testing a broken experience. Then test guest-checkout visibility. A six-week full redesign would mix many changes and make attribution impossible.

2. Design the first A/B test with sample size and decision rules.

Read the worked answer

Test design

Assume the first test is guest-checkout visibility. Baseline checkout conversion is 1.8% (purchase divided by session). A 10% relative lift is 1.98%, an absolute difference of 0.18 percentage points. For 95% confidence and 80% power, use n around 16 x p(1-p) / d^2 per variant:

n around 16 x 0.018 x 0.982 / 0.0018^2 = about 87,000 per variant.

With 40,000 monthly sessions, a 50/50 split gives 20,000 per variant per month, so the test would take about 4.4 months. That is not practical for a six-week window.

Options:

Target a larger, realistic MDE. A 20% relative lift (1.8% to 2.16%, d = 0.0036) needs about 21,800 per variant, or roughly 1.1 months at 40,000 sessions per month. This is feasible.
Restrict to mobile checkout sessions if that is the problem area. This increases relevance but reduces total sample, so recalculate.
Use a sequential or Bayesian method with pre-agreed stopping rules, but do not peek without adjustment.

Decision rule: ship if the treatment lifts checkout conversion by at least 20% relative with 95% confidence and does not increase return rate or fraud. Guardrails: average order value and customer support contacts. Pre-register the primary metric; do not switch to revenue mid-test.

3. What qualitative research should the UX researcher do in five days with eight users, and what should they avoid?

Read the worked answer

Five-day qualitative plan

Recruit eight recent mobile abandoners or mobile shoppers who reached checkout in the last 30 days. Offer a small incentive. Include a mix of guest and account users if possible.

Day 1: write a discussion guide and task script. Focus on the checkout flow, not opinions about the brand. Define tasks: find a product, add to cart, check out as a guest, enter an address, review shipping, and pay.

Days 2-3: run 45-minute moderated remote sessions. Use think-aloud and ask users to share their screen. Observe where they hesitate, what they expect, and how they react to the address error and the account prompt. Record task success, time, errors, and verbatim quotes.

Day 4: run a heuristic review and review session replays of checkout error pages. Compare findings with analytics drop-off points.

Day 5: synthesize by severity and frequency. Produce a short list of fixes with evidence.

Avoid

Do not ask leading questions such as 'Wouldn't guest checkout be easier?' Do not use the eight users to validate a finished redesign; use them to discover problems. Do not treat small-sample percentages as quantitative proof. Avoid testing only the happy path; deliberately test the address error and the hidden guest link. Do not skip recruiting actual mobile abandoners, since colleagues or power users will not represent the target.