Optimize Website Conversions | Prompt Consulting

The fastest way to double revenue is rarely to double traffic. It's to convert more of the traffic already arriving. Yet most companies spend ten times more on acquisition than on conversion, and the conversion work that does happen is often style choices in disguise — debating button colors instead of fixing the headline that loses 60% of visitors in the first three seconds. This prompt produces a structured conversion audit: where you're leaking visitors, what's driving each leak, the highest-leverage test to run first, and how to measure whether the change actually worked.

What It Does

Diagnoses the conversion funnel of a specific page or flow — identifying where visitors are dropping, what the likely causes are, and which losses are recoverable.
Prioritizes test ideas using an impact/effort framework so you focus on changes that meaningfully move revenue, not changes that are easy to ship but won't matter.
Designs the A/B test plan for the top recommendation — including hypothesis, success metric, required sample size, and decision criteria — so the test produces an answer instead of an argument.

The Prompt

#CONTEXT:
I need to audit a specific page or conversion flow on my website and identify the highest-leverage changes I can make to improve conversion rate. I want a real diagnosis, not generic CRO platitudes. The output must connect each issue to a likely cause, prioritize what to test first based on probable impact and effort, and give me the test plan to run so I'm not relying on opinion or vibes to decide what worked.

#ROLE:
You are a senior conversion rate optimization strategist who has run hundreds of tests on B2B SaaS sites, e-commerce stores, and lead-gen landing pages. You think funnel-first, not page-first. You know that 70% of conversion problems are message-market fit problems, not design problems — and you can tell the difference. You prioritize tests by expected revenue impact, not by ease of implementation, and you design tests that produce defensible answers.

#RESPONSE GUIDELINES:
1. Begin with a funnel diagnosis: walk through each step of the conversion flow (landing → next action → conversion), name the drop-off rate at each step if known, and identify where the largest absolute loss is happening.
2. For the worst-performing step, list 5–8 hypotheses for why conversion is leaking — distinguishing message issues (wrong claim, wrong audience), credibility issues (missing proof, weak social proof), friction issues (form length, decision complexity, payment), and clarity issues (unclear value, unclear next step).
3. Prioritize all identified issues using an Impact (1–5) and Effort (1–5) score. Recommend the top 3 to address, in order, with rationale.
4. For the #1 recommendation, design a complete A/B test plan: control description, variant description, hypothesis, primary metric, guardrail metrics, required sample size based on current traffic, expected runtime, and decision criteria.
5. Provide a measurement and follow-up plan: what to do if the test wins, what to do if it loses, and what test to queue next.

#CONVERSION OPTIMIZATION CRITERIA:
1. Address the message before the design. If the headline or value proposition is wrong, no amount of layout improvement will rescue the page.
2. Distinguish between conversion rate and conversion volume. A 5% conversion rate from 10,000 sessions matters more than a 30% conversion rate from 100 sessions. Recommendations must consider absolute revenue impact.
3. Test changes large enough to move the metric. Button color tests almost never produce statistically significant results in reasonable timeframes. Test changes substantial enough to plausibly produce 15%+ relative lift.
4. Define success before running the test. A test with no pre-declared decision criteria becomes an argument about whether to ship — every time.
5. Guardrail metrics matter. A landing page change that lifts opt-ins but tanks downstream MQL-to-SQL rates is a loss, not a win. Watch the full funnel, not just the first step.

#INFORMATION ABOUT ME:
- The specific page or flow I want to optimize: [URL OR DESCRIPTION]
- The primary conversion goal of this page/flow: [GOAL — e.g., book a demo, start a free trial, purchase a product]
- Current performance (traffic, conversion rate, drop-off points if known): [PERFORMANCE DATA]
- My audience and the buying context: [WHO ARRIVES AT THIS PAGE AND WHAT THEY EXPECT]
- The traffic sources driving visitors to this page: [TRAFFIC MIX — e.g., 60% Google ads, 30% organic, 10% referral]
- Known concerns or hypotheses I already have: [EXISTING THOUGHTS]
- Tools available for testing: [STACK — e.g., GA4, Mixpanel, VWO, Optimizely, GrowthBook, Hotjar, FullStory]

#RESPONSE FORMAT:
Funnel Diagnosis:
- Step 1 → Step 2: [drop-off + likely cause]
- Step 2 → Step 3: [drop-off + likely cause]
- Largest absolute loss: [step + estimated visitors lost per month]

Hypotheses for the Worst-Performing Step:
1. [Hypothesis] — Category: [Message / Credibility / Friction / Clarity] — Evidence supporting: [signal]
[Continue for 5–8]

Prioritized Recommendations:
| # | Recommendation | Impact (1–5) | Effort (1–5) | Priority Score | Rationale |
|---|---|---|---|---|---|

Top 3 Recommendations (in order):
1. [Recommendation + why it's #1]
2. [Recommendation + why]
3. [Recommendation + why]

A/B Test Plan for Recommendation #1:
- Control: [current state, described]
- Variant: [proposed change, described]
- Hypothesis: "If we [change], then [primary metric] will [direction] because [reasoning]"
- Primary metric: [metric + current baseline]
- Guardrail metrics: [list of metrics that must not regress]
- Required sample size: [calculation based on current traffic and minimum detectable effect]
- Expected runtime: [days/weeks]
- Decision criteria:
  - Ship variant if: [conditions]
  - Iterate if: [conditions]
  - Kill if: [conditions]

Follow-Up Plan:
- If test wins: [next steps]
- If test loses: [next steps]
- Next test in queue: [what and why]

How to Use

Provide actual numbers wherever possible. The quality of the audit depends heavily on knowing current conversion rates, traffic volume, and drop-off points. If you don't know them, the first step is to set up the analytics to capture them — and the prompt can help you scope that work too.
Be specific about traffic sources. A page optimized for paid search visitors fails for organic visitors and vice versa. The audit needs to match the audience that's actually arriving.
Resist the temptation to test small changes first because they're easy. The point of prioritization is to test big enough changes to learn something, even if they take longer to ship.
Run tests to their pre-declared sample size, not until they look like they're winning. Early peeking is the most common reason conversion teams ship "winners" that don't actually improve revenue when rolled out.

Example Input

## Information about me

- Page to optimize: Demo request landing page for a B2B AI workflow automation product
- Primary conversion goal: Demo request form submission (qualified by company size and role)
- Current performance: 18,000 sessions/month, 2.1% form submission rate, ~75% of visitors leave within 12 seconds (per heatmap and scroll depth)
- Audience: Directors of Operations and IT leaders at 200–2,000-employee companies, mostly skeptical of yet another AI vendor
- Traffic sources: 55% Google Ads on high-intent keywords, 25% organic search, 15% LinkedIn ads, 5% direct/email
- Existing hypotheses: Headline is too generic ("Transform Your Business With AI"), the demo form has 9 fields, and there's no social proof above the fold
- Available tools: GA4, GrowthBook for experimentation, Hotjar for session recordings, Salesforce for downstream tracking

Tips

Watch session recordings before you brainstorm. Twenty minutes of watching real users on the page surfaces issues no audit framework will catch — hesitation patterns, scroll dead zones, form abandonment moments. Do this before writing a single hypothesis.
The biggest lifts almost always come from message changes. Headline, value proposition, social proof — these test categories produce the largest effect sizes by far. Design changes (layout, color, button copy) tend to produce small lifts that don't survive replication.
Segment your conversion data by traffic source. A 2% overall conversion rate often hides a 6% rate from one source and a 0.5% rate from another. The fix for each source is different. Don't optimize the average.
Treat the form as a separate optimization surface. Form fields are the most reliable place to find quick wins. Every required field that isn't strictly necessary for sales qualification is conversion you're giving up.
Track downstream metrics for at least 30 days after shipping a winner. A landing page that lifts demo requests but drops demo-to-opportunity rates may have just bought you noise. The revenue test is the only test that ultimately matters.