Funnel Optimization: Find the Bottleneck Before You Test

A useful funnel review starts with an explanation of where good prospects get stuck. A low submission rate may reflect a confusing form. It may also mean visitors cannot find the service they need, the offer attracts the wrong audience, or the next step asks for more commitment than they are ready to make.

Those problems call for different changes. Use the method below to choose a test, then open the relevant part of the 47-idea library. The library is a reference, not a requirement to redesign everything.

Follow one inquiry through the whole process

Start with the route from the original ad, search result, or referral through the landing page and request form. Continue through confirmation and the staff response. A successful submission is only one handoff in that journey.

Look for evidence of a specific problem. Review where appropriate visitors leave, where errors occur, and what the people handling inquiries repeatedly need to explain. For a medical practice, check whether someone can choose the right office and reach its actual request process. For a software company, check whether the page and sales conversation describe the same product and commitment.

Fix broken links, failed delivery, and unusable controls as defects. Other changes need a hypothesis: what uncertainty or obstacle are you addressing, and why would the proposed change help? Watching someone attempt the task can reveal a problem that a conversion-rate report does not explain.

Choose a test your evidence can support

Prioritize the issue that matters to the business and has a plausible explanation. A clearer description of the service may deserve attention before a button-color experiment when visitors do not understand what is being offered.

If several ideas compete, a simple score can make the discussion concrete. In this worksheet, average impact, confidence, and ease on a 1–10 scale: ICE = (impact + confidence + ease) ÷ 3. The scores are subjective judgments, not predicted lifts.

Record why you assigned those scores. A high number with no supporting observation is a reason to investigate, not permission to ship. The example scores above are invented and do not establish which change your site needs.

Before an A/B test, define the eligible audience, primary metric, smallest worthwhile effect, and decision rules. Estimate the required sample and duration from the traffic that will actually enter the experiment. Statsig’s power-analysis documentation explains the relationship between detectable effect, allocation, and duration.

For a site with limited traffic, a long or insensitive experiment may not answer the question usefully. Observed usability sessions, inquiry review, and checks for broken journeys can support practical improvements without claiming a statistically established conversion gain.

Measure what happens after the conversion

A shorter form might increase submissions while removing information that staff need to qualify a request. Decide which downstream measure would reveal that tradeoff: appropriate inquiries, completed appointments, useful sales conversations, or purchases, depending on the business.

Consider purely hypothetical conversion rates of 2.4% and 7.8%. The second is a 225% relative increase, but those invented numbers establish nothing about inquiry quality, customer value, or whether a real test would produce that result. Keep the numerical calculation separate from evidence about the change.

The same care applies to a series of improvements. Assumed relative gains of 20%, 15%, and 12% compound as 1.20 × 1.15 × 1.12 ≈ 1.546, or 54.6% total lift. That arithmetic assumes each gain persists on the updated baseline. Real tests can lose, be inconclusive, or interact; a sequence of assumed wins does not belong in a forecast as observed growth.

At the readout, record the hypothesis, the result, its uncertainty, and the decision. Keep useful negative findings. If the chosen metric improved while downstream quality worsened, explain both before recommending rollout.

Find the relevant test ideas

Open the stage where you found a problem. These 47 ideas are options to evaluate, not universal recommendations. Use only accurate, authorized customer evidence and actual offer terms. A test should not introduce fabricated urgency, misleading claims, or unnecessary collection of personal information.

Traffic and message match · tests 1–8

• Test 1: The landing page clearly continues the offer promised in the ad
• Test 2: Landing page echoes ad creative (same visual style, colors)
• Test 3: URL structure reflects ad promise (e.g., /demo not /home)
• Test 4: Compare relevant match-type choices against qualified conversions and cost
• Test 5: Negative keywords to filter low-intent traffic
• Test 6: Create separate landing pages for each audience segment
• Test 7: Use dynamic text replacement to personalize headlines
• Test 8: Test relevant location messaging; use customer counts only when you can substantiate them

Landing-page clarity and evidence · tests 9–30

• Test 9: Headline clarity (A: feature-focused vs B: benefit-focused)
• Test 10: Subheadline (remove vs keep vs different messaging)
• Test 11: Hero image (product screenshot vs customer photo vs none)
• Test 12: CTA button position (center vs right vs sticky)
• Test 13: CTA button color (brand color vs contrasting color)
• Test 14: CTA button copy (“Get Started” vs “Start Free Trial” vs “See Demo”)
• Test 15: Social proof placement (above fold vs below fold)
• Test 16: Customer logos (above fold vs below fold vs none)
• Test 17: Testimonial format (text vs video vs photo+quote)
• Test 18: Specific metrics in testimonials, with supporting records and permission
• Test 19: Relevant certifications or awards that the business actually holds
• Test 20: Verified customer counts versus a qualitative description
• Test 21: Whether a real, appropriately disclosed activity indicator helps users
• Test 22: Case study preview (link to full case study)
• Test 23: Copy length (short vs long vs medium)
• Test 24: Benefit bullets (3 vs 5 vs 7)
• Test 25: Explain a real deadline or availability limit clearly, if one exists
• Test 26: Risk reversal (“14-day free trial, no credit card” vs standard)
• Test 27: Pain point emphasis (leading with problem vs solution)
• Test 28: Investigate slow loading on the devices and connections your visitors use
• Test 29: Check layout and interactions across representative phones, browsers, and viewport sizes
• Test 30: Remove navigation/exit points (dedicated landing page vs site page)

Forms and calls to action · tests 31–43

• Test 31: Number of fields (3 vs 5 vs 9)
• Test 32: Single-step vs multi-step form
• Test 33: Field labels (“Email” vs “Work Email” vs “Your Email Address”)
• Test 34: Required vs optional fields (mark optional explicitly)
• Test 35: Inline validation vs submit validation
• Test 36: Progress indicator (multi-step forms)
• Test 37: Pre-filled fields (if returning visitor)
• Test 38: Auto-complete and auto-suggest
• Test 39: Button size (small vs medium vs large)
• Test 40: Button shape (rounded vs sharp corners vs pill)
• Test 41: Button microcopy (“Get Started Free” vs “Start My Free Trial”)
• Test 42: Secondary CTA (“Not ready? Download our guide” below primary)
• Test 43: Chat widget (popup vs embedded vs none)

Confirmation and follow-up · tests 44–46

• Test 44: Clear next steps (“Check your email for access link”)
• Test 45: Set expectations (“Your demo is scheduled for…”)
• Test 46: Collect additional info post-conversion (company size, role, needs)

First useful action · test 47

• Test 47: Immediate call-to-action (“Complete your profile” vs “Invite team” vs “Start first project”)

Leave the review with a decision

Choose one hypothesis, name its owner, and agree on the evidence needed for the next decision. Give the test enough time for the relevant business cycle while keeping the agreed loss and quality limits visible.

The result may be a change to the page, a better inquiry handoff, or a decision to investigate further. Our growth services can help when the team needs someone to connect that analysis with implementation and measurement.