Abandonment wasn't one problem. It was five small ones.

This was a B2B enterprise retail checkout, not a consumer storefront: the buyer was typically a procurement or purchasing agent ordering on behalf of their company, not a person spending their own money on impulse. That distinction mattered for everything downstream. Funnel analytics showed a meaningful share of carts weren't converting to completed orders. What the funnel didn't show, at first, was a single obvious culprit.

Digging into the funnel, drop-off wasn't concentrated at one step, it was spread fairly evenly across the flow. That's specific, useful evidence, not an absence of one: it meant the checkout had several separate small friction points rather than one large blocker. A team that assumed a single root cause would have moved the number a little and declared victory, or fixed the wrong thing and moved nothing.

That reframed the whole approach. Instead of hunting for the one broken step, the candidate list became everything plausible at once: too many steps to complete an order, shipping costs not visible until late in the flow, a form asking for more than it needed, and no visible trust signal for a buyer committing company money to a new vendor.

Four changes went into iterative A/B testing over roughly three to four months: reducing the number of steps, surfacing shipping costs earlier, simplifying the order form, and adding trust badges at the commitment point. These were tested both individually and in combination, since the evenly-spread drop-off pattern suggested no single one would carry the result alone.

Cart abandonment dropped 42%. Conversion rate improved, average order value increased, and checkout completion time went down, all in the direction you'd expect if friction was genuinely the problem.

No single fix was enough on its own. The combination of changes is what moved the number, not any one of them in isolation.

An evenly-spread funnel is itself a data point. Sometimes the honest read is that there isn't one root cause, and the right move is to test several smaller changes.

Targeted changes protect your ability to learn. A rebuild would have shipped a better checkout, probably, but taught us nothing about why.