UX Research Intern in digital banking: triangulating methods to decide at weekly pace at ING Spain

I joined ING Spain's specialized UX Research unit as an intern and integrated into a program that already operated at weekly cadence: 5 unmoderated tests per week, analysis and backlog recommendations every Friday. Over 3 months I executed ~60 usability tests of the Twyp app, the heuristic analysis of Inversión Naranja and contributed to a card sorting and tree testing with 120 participants. I have no post-internship implementation data. This case documents what I executed and the decisions made in that context.
ING Bank España
UX Research Intern
2019 · 3 meses
Madrid, España
Banca digital / Fintech

The problem

Three product teams, one research unit, evidence that couldn't wait for the quarter

ING Spain needed to feed three product teams in parallel with actionable user evidence: Twyp (payments and cashback app), Inversión Naranja (commercial site for investment products) and the main commercial website, which was evaluating a navigation restructure. The UX Research Team operated as an internal service unit with a weekly cadence structured around 5 unmoderated tests per week on UserZoom. I joined this program as an intern and took on part of the execution and analysis.

Discovery

How the discovery was done: what was mine, what was the team's

The program combined four methods in parallel. I explicitly mark what I executed and where I contributed to the analysis within larger team studies.

  • Heuristic analysis of Inversión Naranja (my execution). Expert evaluation of the investment products site: naming problems, visual hierarchy and consistency in the information architecture. Deliverable: heuristic report with prioritized findings.
  • Guerrilla card sorting of Inversión Naranja (my execution). 5 in-person sessions with internal profiles, physical cards. Deliverable: pre-test report for the team’s formal study.
  • Twyp weekly testing program (my execution). 4 thematic sprints on the app’s main flows. ~60 unmoderated tests on UserZoom. Deliverables: 4 sprint closing reports.
  • Formal card sorting, tree testing and SUS (contribution to analysis). Studies executed by the team with my contribution to data collection and analysis.

Insight

Validating only with internal profiles is not a shortcut: it's a blind spot

The guerrilla card sorting with 5 employees from the investment unit yielded 100% success associating fund categories with their descriptions. Read in isolation, the data said «the system works.» When the team expanded the study to 15 participants including profiles without investment products, the rate dropped to 40%.

The gap wasn’t statistical noise. It was proof that the naming problem wasn’t perceptible from the inside. This guerrilla pre-test built the argument that justified the expanded study.

Decisions and tradeoffs

Three methodological decisions within an Intern role

01

Heuristic analysis before user studies

The team was going to run card sorting and tree testing on Inversión Naranja. I proposed doing the expert evaluation first. The alternative was to go directly to user research in open exploration. I chose expert evaluation first.

Tradeoff: sacrificing some open discovery in favor of an initial hypothesis. In practice, the heuristic analysis allowed framing the card sorting questions with specific categories to test rather than in free exploration.

02

Guerrilla card sorting with internal sample as pre-test

Before the team’s formal card sorting, I proposed a guerrilla version with 5 employees from the investment area and physical cards. The alternative was to wait for the formal study with an external sample. I chose to run the guerrilla first.

Tradeoff: biased sample (only internals) with potentially high pre-test value. This pre-test generated the 100% → 40% insight that reframed the problem.

03

Operating within the inherited weekly model, not redesigning it

The 5-tests-per-week pace wasn’t my decision: it was how the bank already ran operational research. The own decision, within that model, was how to structure the weekly report so the product team could act without additional moderation — brief synthesis, actionable findings, prioritized recommendations.

Process

From expert evaluation to the actionable weekly cycle

  1. 01

    Onboarding to the UX Research Team

    Weeks 1–2. Training on tools (UserZoom, report format), shadowing of existing tests.

    Deliverable: operational integration into the research program.

  2. 02

    Heuristic analysis of Inversión Naranja

    Weeks 2–3. Expert evaluation on naming, visual hierarchy and information architecture.

    Deliverable: heuristic report with prioritized findings.

  3. 03

    Guerrilla card sorting of Inversión Naranja

    Weeks 3–4. 5 in-person sessions with internal profiles, physical cards.

    Deliverable: card sorting report (pre-test of the team’s expanded study).

  4. 04

    Twyp weekly testing program

    Weeks 3–14, in parallel. 4 thematic sprints, ~60 unmoderated tests, 4 sprint closing reports.

    Deliverables: Send money Sprint, TC-SMS-PIN Sprint, Discounts 1 Sprint, Discounts 2 Sprint.

  5. 05

    Contribution to expanded team studies

    Weeks 6–12. Support for card sorting with 15 participants, tree testing with 120, SUS program data collection and analysis.

    Deliverables (team’s, with my contribution): tree testing data, continuous SUS analysis.

  6. 06

    Synthesis and handoff

    Final weeks. Consolidation of backlog recommendations and transfer to UX Lead for post-internship continuity.

Results and impact

What the research delivered

01

Naming decision unblocked

The triangulation: heuristic (my execution) + guerrilla card sorting with 100% internal success (my execution) + expanded card sorting with 40% on external profiles (team) built the argument that led the team to redesign fund naming from the perspective of users without investment products.

02

Selective implementation on the commercial site

Tree testing with 120 participants prevented a wholesale redesign. It showed improvements where to implement (Insurance 31%→92%, Shopping 62%→93%) and a regression to leave intact (Twyp 94%→80%). The team recommended implementing only the validated improvements.

03

Risk avoided on Twyp

Weekly tests identified structural friction in the payment flow (39% success rate) and cashback comprehension problems in unbanked users, before reaching production at scale. Without that continuous cadence, the patterns would have gone invisible between quarterly studies.

04

Recommendations adopted to the backlog

The heuristic analysis of Inversión Naranja and the reports from the 4 Twyp sprints were incorporated into the backlog of the corresponding product teams.

What I cannot report

I have no post-internship implementation metrics. Recommendations made it to the backlog, but I cannot honestly report what was implemented, when, or with what measurable impact in production. I was also not the lead for the 120-participant tree testing or the think-out-loud A/B: those were team studies I contributed to in the analysis. The distinction between direct execution and contribution is part of this case.

Reflection

What I learned entering operational research from the outside

This internship was my entry into enterprise-scale research, within a specialized unit that already operated with its own model. The structural lesson was understanding how to triangulate evidence: the heuristic analysis signals what to validate; the card sorting verifies whether users understand the system; the tree testing confirms whether they can navigate successfully. The three together build an argument that no single one alone can support.

The other learning: the value of an intern in a mature operational research program doesn’t come from designing the program — it’s already defined. It comes from reliable execution, timely delivery and the ability to produce actionable syntheses without constant supervision. That’s what I tried to do.

Next case study

From a legacy BMS to a B2B SaaS: design decisions in a project that never reached production

Read case study →