Audit a checkout flow step by step for the exact point shoppers are abandoning at

You are auditing a specific e-commerce checkout flow to find the step most likely responsible for its abandonment rate, using the actual flow and data provided, not a generic checkout best-practices list.

CURRENT CHECKOUT STEPS IN ORDER
Cart review -> create account or guest checkout choice -> shipping address -> shipping method + cost shown -> payment info -> order review -> confirmation.

ABANDONMENT DATA BY STEP (if available)
68% proceed from cart to address, 71% from address to shipping method, only 40% proceed from shipping method to payment.

CUSTOMER TYPE
About 80% first-time buyers arriving from paid social ads.

PAYMENT AND SHIPPING OPTIONS OFFERED
Credit card and PayPal only, no Apple Pay; single flat-rate shipping option, no express tier.

KNOWN CONSTRAINTS THAT CAN'T CHANGE
Built on a hosted e-commerce platform's checkout template — can reorder and hide fields but can't change the underlying page structure.

For each step in CHECKOUT_STEPS, evaluate it against three specific failure patterns rather than generic polish: (1) does it demand information the checkout doesn't actually need to complete this specific purchase (an account creation requirement, a phone number with no stated use, a marketing opt-in presented as if required); (2) does it introduce a cost or fact the customer hasn't seen yet at a point past where they've already invested effort (shipping cost revealed only at the final step, a fee appearing without earlier warning); (3) does it break the customer's expectation of what happens next (a step that looks like it might be the last one but isn't, no visible progress indicator, a back button that clears entered data). If ABANDONMENT_DATA is provided, treat the step with the sharpest drop as the priority even if other steps have more surface-level problems — do not spread recommendations evenly across all steps when the data points at one. If no abandonment data exists, name which single step is statistically most likely to be the leak based on the three failure patterns and say so explicitly, rather than presenting all steps as equally urgent.

For CUSTOMER_TYPE, check whether the flow assumes the wrong default — a first-time buyer forced through account-creation friction, or a returning customer forced to re-enter information the store should already have. Check FIXED_CONSTRAINTS before recommending anything that would require removing them; propose the best available fix that respects the constraint rather than an ideal-world redesign that ignores it.

WHAT NOT TO DO
Do not produce a generic 10-point checkout best-practices checklist (guest checkout, progress bar, trust badges) unless each point is tied back to a specific step in THIS flow with a specific reason it applies here. Do not recommend a full checkout rebuild as the first move; find the single highest-leverage fix first.

OUTPUT FORMAT
1. Step-by-step table: step | failure pattern found (or none) | severity.
2. The single highest-priority step to fix, with the specific reasoning for why it outranks the others.
3. One concrete fix for that step that respects FIXED_CONSTRAINTS.
4. One risk: what could get worse if this fix is made without also addressing anything else flagged.