Guide
Ask Only for Documents That Change a Credit Decision, and Fetch the Rest
Each document request should earn its place by changing a decision. We measure the step document by document, then redesign it so applicants supply only what you cannot fetch.
4-minute read5 November 2026
In Brief
- A funnel that reports one drop between application started and documents submitted hides which request loses applicants, and which uploads are rejected by operations two days later.
- We see four actions: measure each document on its own, test every request against a decision and a rule, fetch records with consent where the rules allow, and watch borrowers on real phones.
- Thinqmesh researches with real borrowers, maps where they drop out, and brings your first redesigned path to a working version within four weeks.
A loan application on a phone usually goes smoothly until the applicant is asked for six months of bank statements, a salary slip and a photograph of a document kept in a drawer at home. Lenders often treat this as an interface question and restyle the upload screen. We see the document step as the place where a lender's internal caution is handed to the borrower one request at a time, and many of those requests have never been tested against the decision they serve.
Measure Each Document, Then Test It Against a Decision
Most funnels report one drop between application started and documents submitted. That hides what matters. For each document, by product and channel, we record how many applicants reach the request, how many uploads fail and why, how many leave and return, and how many are rejected later by operations.
Each request then gets two columns: the credit, fraud or compliance decision it changes, and the rule that requires it. Requests that survive can often be fetched with the applicant's consent, through the central KYC registry, DigiLocker or account aggregators, where your category of lender permits.
Four Actions That Shorten the Document Step
The first two use data you already hold. They show which requests to remove or fetch before any screen is redrawn.
Measure Each Document on Its Own
For each document, by product and channel, count who reaches it, who fails to upload, who leaves, and who is rejected later.
The last figure is usually missing because it sits in a different system. A document that uploads easily and is rejected two days later costs more than one that fails at once, since the applicant has stopped paying attention by then and may have borrowed elsewhere. We bring these counts together in one view.
Get started
- List every document requested for one loan product.
- Ask operations how many submissions of each are later rejected.
Test Every Request Against a Decision and a Rule
Write two columns beside each request: the credit, fraud or compliance decision it changes, and the paragraph of the rule that requires it.
Requests often survive from a paper process or one past loss, and no one has been asked to defend them since. The rules frequently specify less than the checklist does, which leaves the lender room to decide how to obtain the information. A request with no decision and no rule behind it is a candidate for removal.
Get started
- Add the two columns to your document checklist for one product.
- Mark each row that has no decision behind it.
Fetch Before You Ask
Once a request has survived, ask whether the applicant needs to supply it at all.
With the applicant's consent, KYC records can be retrieved from the central registry, documents issued into DigiLocker are treated as equivalent e-documents under the KYC Directions, and account aggregators can supply financial information. Each route depends on your category of lender, so we check the directions that apply to you before designing around them.
Get started
- Identify which directions govern your lending licence.
- List the documents that one of the three routes could replace.
Watch Borrowers Complete the Step on Real Phones
Measurement shows where applicants leave. Watching them shows why.
We run research with real borrowers on their own phones and map the whole path, including the calls they make to a branch when a step is unclear. Each redesigned step is released in a small piece and measured against your earlier figures.
Get started
- Recruit three recent applicants who did not complete.
- Watch each one attempt the document step.
How Thinqmesh Solves This
We begin with research on your own path. Design and development sit together, so what is designed can be built, and each change is measured against your earlier figures. The full engagement runs to eight weeks.
- Week 1
Borrowers, Watched on Their Phones
We watch real borrowers attempt your document step, and read your funnel data beside the findings.
- Week 2
The Path Mapped, Drop-Outs Found
We map the path across the app, the branch and the call centre, and mark where applicants drop out.
- Week 3
The First Path Redesigned and Built
The first redesigned path is built on reusable components from a tested design system.
- Week 4
A Working Version, Measured
Your first redesigned path reaches a working version, with analytics that show whether the change worked.
You finish with a redesigned first path, a map of where applicants dropped out, a design system you keep, and analytics that show whether the change worked. The code is yours.
