Diagnose why a funnel is leaking, not just draw the funnel chart

I have funnel data and know where the biggest drop-off is. What I actually need is a diagnosis of why, ranked by plausibility, so the team investigating it knows where to look first instead of starting from zero.

FUNNEL STEPS AND CONVERSION RATES
Landing page (100%) -> Signup form started (34%) -> Signup form completed (28%) -> Payment info entered (9%) -> Purchase completed (7%)

WHAT EACH STEP ACTUALLY REQUIRES OF THE USER
Signup form asks for name, email, company size, and phone number; payment step reveals the full annual price for the first time.

SEGMENT BREAKDOWN IF AVAILABLE
Mobile users drop from form-completed to payment-info at 12%, desktop users at 41% — a much bigger mobile-specific gap.

ANYTHING RECENTLY CHANGED
The phone number field was made a required field on the signup form three weeks ago; nothing changed on the payment step.

Do this:
1. Identify the step with the worst relative drop-off (not necessarily the worst absolute numbers — a step that loses 80% of a small remaining group can matter less than one that loses 30% of a much larger group, depending on the stated business goal) and justify which one you picked as "worst" against that goal.
2. For that step specifically, generate 3-4 concrete, distinct hypotheses for why users are dropping off there, each grounded in what that step actually requires of the user (a form with too many required fields, a step requiring information the user doesn't have handy yet, a page load or technical friction point, a price or commitment reveal that wasn't expected earlier in the flow).
3. Rank the hypotheses by plausibility given the segment breakdown and recent changes provided — a hypothesis that would predict a specific pattern in the segment data (e.g., "if it's a mobile-specific technical issue, mobile users should convert far worse than desktop at this exact step") should be checked against whether that pattern actually appears.
4. For your top-ranked hypothesis, state the single fastest, cheapest way to confirm or rule it out before committing to a fix.

WHAT NOT TO DO
Do not stop at describing the funnel ("most users drop off between step 3 and step 4") — that much is already visible in the numbers given; the entire value of this exercise is in the ranked, checkable hypotheses about why.

OUTPUT FORMAT
1. Worst step identified, with justification against the stated business goal
2. 3-4 distinct hypotheses for that step, each specific and grounded in what the step requires
3. Ranked plausibility, with reasoning tied to segment/recent-change data where it supports or contradicts a hypothesis
4. Fastest/cheapest way to test the top hypothesis