← All work Alekhya.
Enterprise UX · Data Visualisation

Engine Wash Analytics for GE Aviation

Engine Wash Analytics dashboard

A single source of truth for aviation engine wash performance, so a reliability analyst can open the tool in the morning and know which engines need attention in 30 seconds, not 12 minutes of CSV stitching.

Role
UI / UX Designer, interaction lead
Responsibilities
Stakeholder research Information architecture Wireframing Visual design Usability testing Dev handoff
Project
SAMA Suite (internal GE Aviation + customer pilots)
Timeline
6 months, 2024
Tools
Sketch, Miro, InVision
Context

A tool for people who keep aircraft engines running.

Engine Wash Analytics (EWA) is an enterprise tool built for GE's aviation customers, surfacing recent and historical engine wash performance data. Wash frequency, equipment used, adherence to execution recommendations, engine operational environment, each one has a direct impact on engine durability and fuel burn.

I joined the SAMA design team at GE as the UI / UX lead on this product, working alongside two reliability engineers, a product owner, a tech lead, and four frontend developers in two-week sprints.

EWA home page
The Problem

Analysts were stitching together wash data from multiple sources every single morning.

Reliability analysts and fleet operators were opening four tabs each morning: a vendor PDF, a GE spreadsheet, the operator's own data lake, and an Excel file passed down from the previous shift. The data was all there. Nobody could tell what it meant.

In stakeholder interviews with 8 engineering, analytics, and operations staff, three themes surfaced every single time:

  • Fragmented data, no single version of truth across teams.
  • Interpretation gap, raw values without reference points, "1,240 EGT margin, is that good or bad?"
  • Decision paralysis, every engine looked equally urgent because nothing was prioritised.

The brief, once the themes were in: build a tool a reliability analyst can open in the morning, look at for 30 seconds, and know exactly which engines need attention, and why.

UX process diagram, from discovery through rollout

UX process, discovery through rollout, aligned with the SAMA delivery cadence.

Research

Three quotes shaped the entire product.

I ran semi-structured interviews with the 8 participants, 45 minutes each, and clustered the transcripts in Miro. Six of eight used some version of these three sentences:

"I don't want another dashboard. I want a traffic light that tells me which engines to look at first."Reliability analyst, 9 years at GE
"A number without a benchmark is just noise. 1,240 EGT margin, is that good or bad? I have to go ask someone."Field engineer, customer pilot
"I open four tabs every morning and copy-paste into Excel. If this tool breaks that habit, I will love you."Operations lead, EU region

These weren't stray comments, they were the three pillars the product had to deliver on: triage before detail, context before number, replace the morning spreadsheet.

Design Decisions

Six decisions that shaped the product.

1.

Color scoring over raw numbers.

I ran a sorting exercise with 4 analysts, raw numbers took them 40 to 60 seconds per engine; a red / amber / green chip against each metric's expected range got them to a triage decision in under 5. We sided with scan-ability over precision at the fleet level, and kept the raw number one click away on the detail screen.

2.

Dashboard-first, not search-first.

In 7 of 8 interviews, the first question of the morning was "is anything wrong?", never "tell me about engine X." Starting on a search box would have forced users to already know what to look for. The fleet-wide KPI summary respected the real question.

3.

Scatter plots over line charts.

I tested two prototypes with the same data, one with a smoothed trend line, one with scatter points. Three of four engineers assumed the line meant engine performance was trending in a direction. Washes are discrete events, not a continuous signal, and the line was telling a story that wasn't true. Scatter stopped the false narrative without overclaiming shape.

4.

Tooltips over a documentation tab.

Aviation acronyms (EGT, ESN, HDM, OEI) are dense, and a separate glossary tab was ignored in testing. Inline tooltips, triggered by hovering any acronym, were used three times as often and let users learn the vocabulary while working, not in a separate reading session they never came back to.

5.

Progressive disclosure over a mega-table.

A flat, single-table layout was faster to build and showed everything at once. In testing it overwhelmed users on load, scrolling through 400 rows to find something wrong. Breaking the data into Fleet, Engine, and Wash Event levels, each drillable from the one above, matched how analysts actually asked questions.

6.

On-demand review over push alerts.

Stakeholders worried about alert fatigue in an already-noisy ops environment, every tool was pinging. We shelved push notifications and kept the morning summary as the single trusted entry point. Analysts could opt in later; most never did.

Fleet performance page

Fleet Performance, a KPI summary, ESN-level table, and monthly trend plot in one view.

Engine performance trends page Fuel savings page

Engine performance trends (left) and fuel savings / sustainability (right), drillable from the fleet view.

Raw data export view

Raw Data, each row is a wash event or an engine's current state. Exportable to CSV for the analysts who still live in Excel.

Reflection

Shipped in 6 months, and changed the morning routine for reliability analysts.

After a 6-person usability round (3 engineers, 2 reliability analysts, 1 operations manager):

  • Time to answer "which engines need attention this week?" dropped from ~12 minutes of CSV stitching to under 45 seconds in the dashboard, a ~94% reduction.
  • 5 of 6 participants correctly interpreted the color-scoring system on first exposure, without tooltip help.
  • Qualitative shift: analysts said they'd open EWA as part of their morning routine, a change from the old spreadsheet workflow that only opened reactively.
  • Shipped as part of the SAMA suite, rolled out to internal GE engineering teams and select customer pilots.

The biggest thing I learned: in enterprise analytics, the most important design decision is what you don't show. Every stakeholder asked for more metrics on the home page. The product won because we kept saying no, and kept the first 30 seconds focused on triage.

Limitation: rollout was limited to internal GE teams and a small customer pilot within the project timeline. Longitudinal usage data, "did analysts actually open it every morning three months later?", would have been the honest proof. That measurement sat with the product team after handover.