SETHJACOBS

AgileThought · Client: Deloitte · Fintech · 2019

Making usability debt
cost real money.

Users of a legacy tax platform could only tell me it “took too long.” I used KLM-GOMS to turn that feeling into arithmetic, a projected $109M a year, and the client funded a greenfield redesign of the whole Document Management experience on the strength of it.

Role

Head of User Experience, AgileThought · designed hands-on, led the practice and the vision

Company

AgileThought · Client: Deloitte

Timeline

2019

Status

Delivered to client, 2019

Tools

Sketch, Abstract

Impact

$109M a year, quantified

The short version

  • Anecdotes don't get funded. 6 errors a week × 7 minutes × $30/hr × 100,000 accountants ≈ $109M a year does.
  • I let the number set the priorities: task automation, shareable views, a version ledger, tagging at upload.
  • I tested every two weeks with all three personas, so launch was a non-event.

01 · The problem

The complaint you couldn't put a price on

I inherited a legacy Document Management system that was slow and missing features, and nobody could quantify the pain. Users told me tasks took too long.

Document Management: list view with metadata, tax year, legal entity, and the calendar sidebar

The redesigned Document Management workspace: the document list with its metadata, tax year and legal entity, and the calendar sidebar showing next due dates and workflow status.

Grid, calendar and workflow share one screen, which is the whole point. The old system made those three separate trips.

02 · Who I studied

Three roles, one complaint

THREE PERSONAS, INTERNATIONAL POOL, WEEKLY CADENCESAStaff AccountantMAJORITY OF USERS~70%“It takes too long. I can't tell you why. It justdoes.”GOALSMove documents through without re-workFind the right version first timeFinish the season without heroicsFRUSTRATIONSRepetitive hand-offs tracked by memoryRe-configuring the same view every sessionTMTax ManagerREVIEWS AND DELEGATES“I need to know it's done, not chase it.”GOALSDelegate with confidenceSee completion without askingSign off cleanlyFRUSTRATIONSAmbiguity about whether work finishedSigning is cumbersomeTATax AnalystINSIGHTS AND CLIENT DATA“Half my job is finding the version everyone agreedon.”GOALSTrace a document's historyShare working context with the teamFRUSTRATIONSNo clear record of who changed whatEvery persona tested every two weeks through the engagement, which is why launch surfaced no major usability issues.
An international pool, tested every two weeks for the length of the engagement. The Staff Accountant's line is the one the whole business case grew out of.

03 · The method

Turning a feeling into arithmetic

You don't get a redesign funded off a hallway grievance, so the first job was turning the complaint into something with a number attached. Discovery paired persona and journey work with a System 1 / System 2 read of where the interface was demanding thought it had no right to demand, and KLM-GOMS predicted the task and error-recovery times underneath that. What came out the other end was a defensible nine-figure loss projection. The conversation changed the same week.

04 · The arithmetic

Four numbers the client couldn't argue with

TURNING “IT TAKES TOO LONG” INTO A NUMBER6errors / weekMEASURED IN TEST×7min recoveryKLM-GOMS MODEL×$30per hourLOADED COST×100,000accountantsCLIENT SCALE=$109.2M / yearTHE COST OF THE INTERFACE THEY ALREADY HAD
Every term in it has a source. We counted the errors in testing, I modeled recovery time with KLM-GOMS, and the loaded cost and headcount came from the client's own figures. That mattered more than the total did, because anyone who wanted to argue with the conclusion had to argue with one of the four numbers instead.

05 · The journey

Where a delegation breaks

JOURNEY MAP · THE DELEGATION SCENARIOTax Manager delegating a workpaper package01Define02Specify03Sign04DelegateFEELINGDOINGDecides what has to go out and towhom.Selects workpapers by hand, one at atime.Prints, signs, scans, re-uploads.Hands off, then chases status byemail.FRICTIONManual workpaper selection, easy tomiss one.Signing is the single most cumbersomestep.No visibility into completion;risk-checking overhead.WHAT I DESIGNEDAutomate the selection from the task.Digital signature in the flow.Automated hand-off states and atransparent version ledger.
Mapping a single scenario (a manager delegating a workpaper package) was enough to locate the friction well enough to price it. Signing was the deepest dip on the map, which is why digital signature led the build rather than something more visible.

06 · The forensics

The research that built the case

User Personas

User Personas

Three personas were identified: Tax Manager (handles tax documents and reviews work), Staff Accountant (majority of users, handles tax documents), and Tax Analyst (provides insights and data on clients). Users were international and involved throughout the entire process on a weekly cadence.

KLM-GOMS & The Formula

KLM-GOMS & The Formula

Applied the KLM-GOMS HCI method to predict task completion and error recovery time. The formula: (# of errors) × (avg. recovery time) × (employee cost) × (# of employees) = monetary losses. Result: 6 errors/week × 7 mins @ $30/hour × 100,000 accountants = $2,100,000/week or ~$109,200,000/year in losses.

Journey Mapping

Journey Mapping

A journey map for the Sr. Manager delegation scenario was created, covering Define → Specify → Sign → Delegate stages. Key friction points surfaced: cumbersome signing, ambiguity about completion, and risk-checking overhead. Opportunities identified: automate workpaper selection and enable digital signatures.

(# of errors) × (avg. recovery time) × (employee cost) × (# of employees) = the cost of bad UX.
The formula that reframed the project

07 · The redesign

What the number bought

The figure did more than fund the work. It decided the order of it. Each stream went after a cost we had measured. Automation ended the manual hand-off tracking, shared grid views closed the knowledge gaps between roles, a version ledger settled who changed what, and tagging at upload collapsed two steps into one.

08 · The solution

The redesign, screen by screen

Task Management: document grid with the Looking Ahead calendar and task automation states
View Customisation: a custom grid with related tasks, sharing and status columns

Task management, with the Looking Ahead calendar and task automation states, and in front of it view customisation: a grid with related tasks, sharing and status columns alongside the calendar and workflow panel.

Task states are explicit and automated, so nobody is tracking a hand-off from memory any more. Views save and share, so a colleague inherits your working context instead of rebuilding it.

Version History: a vertical version ledger beside a birds-eye view of commenting
Upload Tagging: bulk upload with duplicate handling and a minimisable status bar

Version history as a vertical ledger with timestamped contributors, beside a birds-eye view of commenting in the spreadsheet itself. Then bulk upload, with duplicate handling, task association, and a footer status bar that can be minimised while you carry on.

Comments travel with the document, even outside the repository. And tagging happens at upload, which retires the old upload-then-go-back-and-tag routine.

09 · Reflections

What I'd carry forward

  • My KLM-GOMS analysis priced the usability problems at ~$109M a year. One number set scope, priorities, and budget.
  • Task automation ended manual hand-off tracking across all three personas.
  • Shareable grid views let people hand over working context instead of rebuilding it.
  • I tested biweekly with an international pool, so launch was a non-event: no major issues post-release.