Raindrop: making pension search feel possible

Raindrop helps people find and combine old pensions. The service was genuinely useful, but the form made the work feel harder than it was. Only about one in eight people reached the end. On mobile, small controls, late validation and no safe way to pause turned a sensitive financial task into a test of patience.

My job was not to make the form look newer. It was to help people understand what Raindrop needed, why each detail mattered, and how much work was left, while still collecting information the operations team could act on. I worked inside real limits: a six-week window, Raindrop's existing Gaudi platform, and an FCA-regulated transfer context where the wrong nudge could cost someone their employer contributions.

The completion rate was only the final symptom. The recordings showed people leaving before they had given Raindrop anything at all.

• Two in three people left within thirty seconds, before entering a single detail.
• Among those who did submit, roughly four in ten produced formatting errors that only surfaced at the end.
• Mobile sessions showed about three times more rage-clicking than desktop.

This was not a long-form problem, it was an uncertainty problem. People did not know what they would need, why some details mattered, or whether they could come back later without losing their work. The old journey made it worse: it treated National Insurance as a step of its own, asked for a password and investment consent at the final hurdle, and showed a low chance of finding a pension without helping anyone improve it.

Two decisions did most of the work. First, the flow explains that more detail improves the chance of a match, and it warns people not to transfer a current workplace pension if that could cost them employer contributions. These messages sit next to the decision they affect, where they prevent a mistake instead of becoming footer copy nobody reads.

Second, I separated consent from consequences. The old final screen bundled account creation, terms and default-fund consent into one moment. The new Combine step explains the plan and its 0.75% annual charge before anyone agrees, the signature step says why providers need a signature, and a Review screen lets people check everything before it is submitted.

I did not shrink the desktop form into a narrow column. Desktop keeps the whole journey in view; mobile converts the same structure into a vertical progress rail with larger controls, shorter lines and actions placed after the content they act on. The named journey, the Save action, the guidance and the validation logic stay identical. Only the presentation changes.

The ending matters as much as the entry. The final flow runs from the first detail through pension entry, plan choice and signature to a Summary and Review, then a specific success state that confirms the searches were submitted and tells people what to expect. A financial form should never leave someone wondering whether anything happened.