What to Fix First on Your Website: A Practical Guide for Small Teams

Your team has time for one website improvement: replace the homepage image, simplify the gallery, or make sizing help easier to find. Choose the change with the clearest evidence that it addresses an obstacle to a task visitors need to finish.

To prioritize website improvements, document the obstacle, check comparable visits, and assess whether people recover and complete the task. Then decide whether to change the page, investigate further, or defer the idea.

The worked example and copyable worksheet below take that decision through to a follow-up review.

Start with a task someone needs to finish

Name a task that matters to visitors and your business. For a shop, it might be choosing an available size and adding a product to the cart. For a service business, it could be checking the service area and submitting an enquiry.

Define its observable start and finish. GOV.UK’s completion-rate guidance uses explicit transaction boundaries. Applying that principle here keeps the review consistent.

For the shop example, start with interaction with the size selector and finish with a confirmed cart addition. Opening the selector does not prove an intention to buy. Adding to the cart completes this intermediate task, not a purchase.

Describe the sequence before explaining it

Illustrative example, not customer data. The sizing scenario and decisions throughout this guide are hypothetical.

A mobile visitor opens a product’s size selector, scrolls through the information, returns to the selector, then leaves the page. No cart addition is visible.

“Sizing help may be difficult to find” is one interpretation. The required size could be unavailable, or the visitor could be comparing products or interrupted.

GOV.UK’s research-analysis guidance separates what researchers saw or heard from what they think it means. Apply that discipline here: save the observed sequence alongside its session reference before proposing a fix.

Review the surrounding actions in Revisit

Use Revisit’s session replay to review the recorded visit:

1. Adjust playback speed and use Skip idle to reach the relevant part. Turn Skip idle off when examining a pause.
2. If a “What happened” entry is available, select its timestamp and watch the surrounding actions.
3. Note what appeared after the interaction and whether the task finished later.

The Summary panel can contain “Worth fixing” suggestions and an interest score. Treat them as prompts to inspect the actions, rather than a business-priority ranking. Some sessions have no AI summary.

Check comparable visits, including completed attempts

Review visits that reached the same step on the same device category and page version. Include completed attempts to see whether the sequence also occurs when visitors succeed.

Record dates and how you selected visits. A sample chosen for repeated clicks differs from a sample of visits that interacted with the size selector.

Revisit’s heatmaps offer device and date selection. From an element ranking, you can open sessions that clicked that element. The lookup carries the device, dates, and a click-type filter. Review comparable visits beyond the flagged group too.

Count sessions reviewed and sessions containing the sequence. Repeated taps within one session do not represent additional affected visitors.

Keep these counts tied to the sample. Nielsen Norman Group’s sample-size guidance distinguishes qualitative learning from reliable numerical measurement. Applying that distinction, a sample selected for difficulty cannot establish how common the problem is across all traffic.

Stop reviewing when the next action is clear

There is no universal recording count for choosing a fix. Let the unresolved question direct your next review.

If recovery is unclear, follow the later actions. If the sequence also appears in completed visits, compare what happens next. If recordings cannot reveal what information people needed, watch a likely user attempt the task.

For a modest change, document the obstacle, its effect on the task, and why the proposed remedy addresses it. If the explanation remains weak, choose an investigation. A wider redesign needs broader evidence.

Separate task outcomes from obstacles

Record the outcome as complete, no completion observed, or unknown. Separately note any obstacle and whether the visitor recovered:

• Mark completion when the defined finish is visible, including any difficulty overcome.
• Record no completion observed when the visitor leaves before that finish. Keep the reason open.
• Mark the outcome unknown when the recording does not show enough to judge.
• Describe a visible blocker when it prevents the next required action. Check for a later workaround before concluding the task stopped there.

Nielsen Norman Group’s severity guidance considers frequency, impact, and persistence. Use those factors to distinguish recoverable difficulty from a confirmed obstacle that stops the task. An uncommon severe blocker can still warrant prompt action.

Our guide to rage clicks and dead clicks explains those signals. Inspect what follows the clicks before assigning severity.

Compare candidates and choose the next action

Use three possible decisions: change the page when the obstacle and remedy are supported; investigate when the explanation is uncertain; defer when the review has not established a meaningful problem.

Illustrative example, not customer data. Observations and effort assessments are hypothetical.

Candidate	Evidence and next action
Move sizing help	The selector, scroll, and return sequence recurs in comparable visits; some show no cart addition. The reason is uncertain. Placement looks easy to change, but investigate visibility first.
Simplify the gallery	Repeated image browsing appears in visits that reach cart addition. No clear obstacle is established. Redesign takes more work; defer.
Replace the homepage image	A teammate prefers another photograph. No obstacle to the chosen task is documented. The edit is easy; defer within this review.

The sizing issue earns the next investigation. It does not yet justify moving the link.

Continue the fictional example: inspection finds the guide below a long description. During a task observation, a likely shopper searches for measurements without finding it. The guide contains relevant measurements, and the required size is available. This supports trying a clearly labelled “Size guide” link beside the selector.

Baymard’s apparel usability guidance recommends that placement. The research supports the candidate; observations on your page establish its relevance. Moving a link will not resolve missing stock or inaccurate measurements.

Nielsen Norman Group’s prioritization guidance uses criteria such as user impact and feasibility. Here, use effort to break a tie between meaningful, supported improvements. Ask the person making the change to consider implementation, checking, and reversal. This is practical editorial guidance, not a validated scoring formula.

Write the decision before changing the page

State what the change should help someone do. In the example, visitors needing measurements should find the guide, return to the selector, and continue. More guide clicks alone would not demonstrate success. Some visitors already know their size.

Copy this decision worksheet

This is a team note, not a built-in Revisit feature. Replace the illustrative entries with your evidence; leave unknowns explicit.

Illustrative example, not customer data.

Field	Working record
Task	Choose an available size and add the product to the cart.
Observation	Open selector, scroll, return; no cart addition visible before leaving.
References	Add session references, relevant moments, and task-observation notes.
Review scope	Record dates, page version, device, selection method, and number reviewed.
Recurrence	Count affected sessions separately from repeated taps.
Outcomes	Record completion, no observed completion, or unknown; note blockers and recovery separately.
Explanation	Sizing help may be hard to find. Check stock and information quality.
Decision	After the visibility check supports it, move the Size guide link beside the selector. Defer other candidates.
Expected effect	People needing measurements can find the guide, return, and continue.
Delivery	Assign an owner; record release date and how to reverse the change.
Review	Set a checkpoint appropriate to traffic; a date does not guarantee enough evidence.
Follow-up	Record evidence, uncertainty, and whether to keep, revise, reverse, or investigate.

Check the same task after the change

Review comparable mobile visits after release. Look for the original sequence and follow subsequent actions. Check whether visitors find the guide and continue, and whether a new obstacle appears, such as difficulty closing it.

Keep task boundaries unchanged. With configured funnel steps, Revisit supports comparison periods, completion totals, and visitor or session counting. Keep the counting mode consistent and show both entrants and completions alongside any rate.

Record other changes, including promotions, stock, or traffic sources. Nielsen Norman Group’s guidance on confounding variables explains why other conditions can influence outcomes. A before-and-after difference alone does not establish that your edit caused it.

Decide what to do with the evidence:

• Keep the change provisionally when observations support easier completion and reveal no new obstacle.
• Revise or reverse it when observed attempts show a new problem or the remedy misses the original difficulty.
• Investigate further when relevant attempts are sparse or observations conflict. An unchanged rate alone does not settle the decision.

Fewer clicks or shorter visits are insufficient. Record what visitors could finish. If the next obstacle appears during payment, use our guide to investigating checkout abandonment.

When traffic is low, observe a task directly

If an A/B test cannot answer your question within a useful timeframe, investigate the suspected obstacle directly.

Ask actual or likely users: “Choose this product in the size you would buy and add it to the cart.” Do not tell them to open the guide or reveal its location.

Watch before helping and record any assistance. Afterwards, ask what information they needed where they paused. This applies GOV.UK’s moderated usability-testing method, which uses defined tasks without instructions that reveal the solution.

Use these attempts to assess the obstacle and proposed change, without treating them as a site-wide conversion estimate. If the guide becomes findable but its measurements remain unclear, record that as the next problem to examine.

For your next review, choose one task and complete the worksheet through the decision row. Assign the next action, even if that action is a focused observation before any edit.

Use your visitors’ sessions to choose your next website fix.