Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Conversion Rate Optimisation › Beginner

Capstone: Auditing a Landing Page and Proposing Tests

The capstone brings the beginner level together. You score an invented worktop company's landing page with the audit scorecard and checklist, separate the blockers to fix from the ideas worth testing, support each finding with evidence from more than one source, and finish with a ranked set of hypotheses another practitioner could review and build.
Capstone projectBeginner levelModule 3: Research and first hypotheses7 minUpdated

Brief

Graniteway Worktops is a fictional family business that makes and fits stone and quartz kitchen worktops across one region. It sells through a free measuring visit: a fitter comes to the kitchen, templates the worktops and the business sends a fixed price. Almost every sale starts with a request for that visit.

The business runs search ads and a little paid social to a single landing page. Traffic has been steady for months, but requests have not kept pace, and the owner has been offered a redesign by a web agency. Before paying for one, the owner wants an independent audit: what is wrong with the page, how sure we can be, what should be fixed at once and what should be tested, and in what order.

Your role is the conversion specialist. A colleague will act as reviewer and the owner will approve the work. The deliverable is an audit pack, not a redesign: findings with evidence, a list of fixes and a ranked set of hypotheses ready for the backlog.

Inputs and constraints

Everything below comes from the owner and the web developer. Treat it as your evidence, and say where you would want more.

Illustrative example — not a real result

Graniteway Worktops: the page, the figures and the research

SourceWhat it says
Search ad (most served)"Quartz Worktops Made & Fitted · Free Home Measure · Fixed Price Quote"
Paid social adPhoto of a finished kitchen; "Love your kitchen again. Book a free measure"
First screen on a phoneSite menu; headline "Crafted in stone since 1998"; a slideshow of kitchens; no button visible
Further downMaterial descriptions (granite, quartz, marble) with technical specifications; a gallery; an "Our story" section
FormNear the bottom: eleven fields including kitchen dimensions in millimetres, preferred stone and budget; button "Send"
ProofThree unattributed quotes in the footer; no reviews, no guarantee shown, although a written guarantee exists
Measure (last 90 days)Figure
Landing page visits6,300 (search 4,100, social 2,200)
Visits on a phone4,400
Measure requests76, so a conversion rate of 76 ÷ 6,300 ≈ 1.2%
Search visitors' rate63 ÷ 4,100 ≈ 1.5%
Social visitors' rate13 ÷ 2,200 ≈ 0.6%
Visitors reaching the form940
Form starts / submissions230 / 76
Requests recorded in the inbox but not in analytics19 in the last month (the thank-you message has no URL of its own)
ResearchFinding
Recordings: 40 phone sessions24 swiped the slideshow, then left without scrolling; 9 typed in the dimensions field, cleared it and left
On-page poll on exit (31 answers)12 "what does it cost roughly?"; 7 "do I need to measure myself?"; 5 "how long until fitted?"
Sales call notesCallers often ask whether the measure visit is really free and whether there is a guarantee
Speed (field data, phones)Largest element shown after 4.9 seconds; the slideshow images are full desktop size

Constraints you must respect:

  • The developer has a few days of time this quarter; nothing that needs a rebuild of the site.
  • Prices are set by the owner and will not change; the page can only present them more clearly.
  • Traffic is modest, so only a small number of tests can run in a quarter.
  • The owner wants a plain-language summary, not a spreadsheet.

Task

Work through the steps in order. Each produces part of the audit pack.

  1. Score the page. Rate every row of the landing page audit scorecard as missing, partial or solid, separately for search and social visitors where their promises differ. Score the two gates, message match and tracking, first, and state the band the page falls in.

  2. Run the checklist. Go through the landing page checklist on a phone with each ad beside the page, and note every line that fails. Use the failures to check that your scorecard missed nothing.

  3. Read the funnel. Lay out visits, form reached, form started and submitted as a Funnel, by device and by source where you can. Say where the largest absolute loss is and which scorecard rows it points to. Remember the tracking gap before trusting any rate.

  4. Attach evidence to each finding. For every weak row, list the sources that support it, using the method groups in the guide to conversion research methods. Mark findings that rest on a single source and say what research would confirm them.

  5. Sort blockers from hypotheses. Blockers are things that stop a willing visitor converting or stop you measuring: fix them directly and list them as tickets. Everything else where reasonable people could disagree becomes a candidate hypothesis.

  6. Write the hypotheses. For each candidate, follow the tutorial on writing a test hypothesis and fill the core fields of the experiment hypothesis template: evidence, problem, change, hypothesis sentence, primary metric, guardrail and decision rule. One change per hypothesis.

  7. Rank them. Order the hypotheses by the traffic each affects, the size of the loss it addresses, the strength of its evidence and the effort it needs. Say which could run as an A/B Test with this traffic and which are better shipped as considered changes and watched.

  8. Summarise for the owner. In plain language, explain the band, the fixes for this week, the first test and whether the redesign is needed now.

Illustrative example — not a real result

Checking your reading of the funnel

Visits 6,300 → form reached 940: a loss of 5,360, by far the largest. Form reached 940 → started 230 → submitted 76: losses of 710 and 154.

The first screen and the length of the page above the form matter more than the form itself. But the 19 unrecorded requests in one month mean the true CVR is higher than 1.2%, and every rate here is understated until the thank-you state is tracked. Social visitors convert at well under half the search rate, which fits a first screen that answers neither ad.

Your figures may differ if you make different, stated assumptions. What matters is that each step is shown and that tracking is fixed before any rate is used to judge a change.

Evaluation criteria

Review your own pack against the table, then swap with someone else if you can.

CriterionWhat a strong answer showsCommon gap
GatesMessage match and tracking scored first; the tracking gap treated as a blocker; band statedThe tracking gap noticed but rates still compared as if accurate
ScorecardEvery row rated with a reason; search and social scored separatelyOne score for both sources, so the social mismatch is hidden
Checklist cross-checkFailed lines listed and reconciled with the scorecardChecklist skipped, or run on a desktop
FunnelLargest absolute loss located before the form and tied to the first screenEffort spent on the form because it is the most visible problem
EvidenceEach finding backed by named sources; single-source findings flagged with the research that would confirm themRecordings quoted as proportions, or a poll treated as representative
Blockers versus hypothesesTracking, slow images and the missing button fixed directly; uncertain changes kept for testingTesting the slow slideshow or the untracked thank-you state
HypothesesTemplate fields complete; one change each; a mechanism; a primary metric and guardrail; a decision rule written in advance"Redesign the page" as a hypothesis, or no guardrail on lead quality
RankingOrder justified by traffic, loss, evidence and effort; honest about what this traffic can testRanked by preference, or more tests planned than the traffic supports
Owner summaryPlain words, a clear first week, a view on the redesignJargon, or a list with no recommendation

Submission checklist

Before you call the pack finished, confirm each item:

  • The gates are scored first and the band is stated.
  • Every scorecard row has a rating and a reason, for each traffic source.
  • Failed checklist lines are listed and match the scorecard.
  • The funnel is laid out with the largest loss named.
  • Every finding lists its sources, and single-source findings are flagged.
  • Blockers are written as tickets with the fix.
  • Each hypothesis has every core template field filled and one change only.
  • The ranking is explained and says which hypotheses can be tested with this traffic.
  • The owner summary fits on one page and answers the redesign question.

When the pack is complete, mark the lesson as done on the beginner conversion optimisation level page. If a criterion exposed a gap, the quickest revision is the matching exercise: annotating a landing page, simplifying a quote form or writing hypotheses from research notes.

Pages this lesson applies

Templates and checklists used

Terms used on this page

Also related

Related