Bank of America's Support Services division operated three internal tools used daily by teams across 18 lines of business. The tools handled print request management, inventory operations, and cash logistics — critical infrastructure with no formal UX oversight since launch.
Leadership wanted to understand why the tools required so much informal support, and whether they should be replaced, rebuilt, or redesigned.
Three months of discovery before a single wireframe. The goal was to understand the actual operating cost of the tools, not just the reported experience of the people using them.
I designed the research to surface problems that users had normalized. Legacy tool users adapt to failure over time; self-reported satisfaction scores reflect that adaptation, not the tool's actual usability.
By pairing contextual observation with heuristic analysis, the research could separate user perception from product reality, and give leadership the kind of evidence needed to make a build-or-buy decision.
Executive sponsors, IT architects, and business ops leads — to map system ownership, identify undocumented constraints, and establish success criteria.
Observed users in their actual work environment. Shadow sessions with four roles revealed workarounds invisible in any survey.
10-dimension perception study using a 1–5 scale. Results were deliberately read against the heuristic.
Heuristic evaluations were conducted as a standard quality lens across all products.
Each persona represents a distinct relationship with the tools, and a distinct failure mode. Together, they revealed a system held together by institutional knowledge rather than usability.
Acts as the informal help desk for the entire team. His deep tool knowledge masks the system's lack of documentation; new users can't get started without finding him first.
Spends 2–4 hours daily transferring data manually between QDirect and SSOS. A gap in the technology stack, not a user error, consumes 25% of her working hours every day.
Cannot complete basic tasks without asking for help. Onboarding relies entirely on institutional knowledge that lives in people, not the system, creating a recurring failure point for every new hire.
Discovers machine outages through the production calendar breaking, not through the system alerting him. Reactive workflows cost the team hours of recovery time that proactive alerts would eliminate.
The survey measured usability perception across 10 dimensions on a 1–5 scale. The results painted a deceptively positive picture, with high scores on confidence and ease of use, until stacked against the heuristic findings.
Low scores on “System is not complex,” “System is consistent,” and “Not difficult to use” were the honest signal. Everything else reflected years of accommodation, not actual usability.
Where the survey said users were comfortable, the heuristic said the product was failing them. Four critical issues, six improvement areas, one success.
| Principle | Description | Pass | Rating | Finding |
|---|---|---|---|---|
| Visibility of status | Interface keeps users informed through appropriate feedback. | ✓ | Improve | Labeling hierarchy, spacing and size can be increased. |
| Linguistic clarity | Interface uses user language, not internal jargon. | ✓ | Improve | Descriptions of printing terms needed for unfamiliar users. |
| User control | Clearly marked action items; users can avoid getting stuck. | ✓ | Improve | Save as Draft, Submit, and Cancel need to stand out more clearly. |
| Consistency & standards | Follows internal and industry conventions; minimal cognitive load. | ✗ | Critical | Interface is not within current brand standards or UI patterns. |
| Error prevention | System checks for error-prone conditions before user commits. | ✓ | Improve | Some areas not providing preventive notifications at point of input. |
| Recognition over recall | User should not need to remember info across the interface. | ✓ | Improve | Navigation does not always prominently highlight current location. |
| Flexibility & efficiency | Design caters to both inexperienced and experienced users. | ✗ | Critical | Printing terminology needs descriptions; new users cannot understand requirements. |
| Aesthetic integrity | Interface does not contain irrelevant or rarely-needed information. | ✗ | Critical | Requestor and Admin roles need separate screens; Admin visible during requestor flows. |
| Helps users | Error messages in plain language, precise and constructive. | ✓ | Success | Error messaging is clear and constructive — a genuine strength. |
| User support | Help documentation is easy to locate and search. | ✗ | Critical | No assistance aids can be located anywhere in the system. |
Mapping user journeys, frontstage actions, backstage processes, and supporting systems across six phases — revealing where the workflow broke down and where future-state improvements would have the most impact.
| Phase | 1Request intakeLOB submits job | 2Setup & transferData moves to systems | 3Print productionFiles released & run | 4FulfillmentPrint → kit → mail | 5ReconciliationCount verified & billed | 6Reporting & closureJob closed, reports issued |
|---|---|---|---|---|---|---|
| Evidence | SSOS request form LOB job ticket | Email confirmation File count notification | QDirect job queue Print schedule | Printed materials Kitting & mailing items | Recon. email to LOB Billing worksheet (Excel) | Report download Charge-back record |
| Customer journeyLOB / requestor | Submits print or fulfillment job via SSOS form — or daily automated batch triggers wizard flow needed | Receives email with file count — only available status update no in-app visibility | No access to production status or print queue zero visibility | Receives finished printed, kitted, or mailed output | Receives reconciliation email and charge-back to LOB cost center manual billing process | Must request reports from staff — cannot self-serve no self-service reporting |
| Line of interaction | ||||||
| FrontstageSSOS interface | SSOS displays request form — admin and requestor views combined on same screen role conflict | Email notification sent to team with file count email only — no in-app status | No connection to QDirect; production data fully siloed from SSOS data silo | No fulfillment tracking in SSOS; production calendar absent manual workaround | Billing worksheet generated via manual Excel manipulation — 2–4 hrs daily critical time cost | 15 report types available — all manually generated and pulled by staff no email / download option |
| TechnologySystems in use | SSOS (Classic ASP · MS SQL 2012) 30 users · 4 bank sites | QDirect → NotePad / Excel → SSOS no automation between systemsCashBag → Tecsys (1× daily manual) | QDirect print control center 56 users · 1MM units/year | Tecsys WMS + The Vault (separate) 11 users · 60K sq ft warehouse | SSOS billing worksheet + Excel CashBag AIT (~$5,500/mo) | SSOS + Tecsys report writer both require manual staff pull |
| Line of visibility | ||||||
| BackstageOperations · Print Fulfillment · Inventory | Request info stored in QDirect; partitions created; super users train new staff — no in-app guidance knowledge dependency | Data extracted from QDirect via NotePad & Excel, then manually loaded into SSOS. Inventory team enters new items into Tecsys AND The Vault separately — up to 30 new items per day duplicate manual entry | Files released in QDirect; print team reviews and queues jobs (2K–10K prints/day, up to 100K). Time entry for 30 team members entered manually manual time tracking | Print → kitting → mailing across sites. Staff emails each morning to check which sites can print if system is down. Tecsys and The Vault checked separately for stock production calendar needed | Team reconciles print job count against SSOS data. Fulfillment manually manipulates data monthly for CashBag charge-backs to LOBs 15 manual reconciliation projects | Staff pulls LOB reports from Tecsys on request. Ops manager tracks data across 15 separate Excel workbooks no delegation possible |
| Line of internal interaction | ||||||
| Support processesArchitecture · Compliance Vendor management | SSOS on Classic ASP — not compliant with ADA/WCAG 2.1 or MRF standards compliance risk · Rec: Refresh → ASP.NET | No automated integration between any systems. CashBag and SSOS both on MS SQL Server 2012 — extended support ended 2022 tech debt · Rec: automate all data feeds | QDirect on CentOS OS — not supported by bank monitoring tools. Confidential print data not encrypted in transit or at rest. Only 1 external support person globally security gap · Rec: Buy → v6.0 + IPP | Tecsys vendor relationship difficult — slow defect response. Tecsys–Vault integration incomplete; full encryption not possible. Keyboard navigation broken (WCAG gap) Rec: Update or Buy; engage vendor mgmt | No ARM-managed access controls in QDirect. BTConverge data flows unregistered in BTConverge registry access risk · Rec: consolidate CashBag into SSOS | Tecsys report writer capability undertrained; SSOS reporting architecture on legacy stack — hard to extend Rec: expand in-app reporting options |
| Recommended future state | ||||||
| Redesign recommendations | Request Wizard: step-by-step guided form with pre-filled profiles for returning users. Role-split views: Requestor vs Admin separated. Migrate to ASP.NET; achieve full WCAG 2.1 A/AA compliance | Automate QDirect → SSOS data feed; eliminate NotePad/Excel handoff entirely. Decommission CashBag AIT; consolidate into SSOS (save ~$5,500/mo). Automate Tecsys ↔ Vault inventory sync | Production scheduling dashboard in SSOS hub pulling live QDirect data. Automated service alerts on equipment failure. QDirect upgrade to v6.0 — web client, SSO flow, WCAG 2.0 compliance | Production calendar visible to all staff — no more morning emails. Tecsys customized user views to remove unused features. In-app + email notifications with user preference settings | Automated billing reconciliation — eliminate manual Excel charge-back. ARM-managed access controls for QDirect. CashBag data surfaced inline in SSOS dashboard | LOBs can self-pull reports from SSOS hub: email, download, and on-screen view. Centralized Support Services Hub delivers full end-to-end visibility for all roles and LOBs |
Beth was spending 2–4 hours every day on manual data transfers between QDirect and SSOS, roughly 25% of her working hours consumed by a gap in the technology stack, not a user error or training issue.
New users couldn't complete basic tasks without finding a super user. Institutional knowledge wasn't documented anywhere; it lived in specific people. Bradley's role had partially become being the person others called when stuck.
The survey data looked acceptable because the people answering had been using the system long enough to adapt to it completely. Their sense of normal was calibrated to a broken baseline.
Where technical debt is too deep and vendor relationships are untenable. The cost of building on the existing stack exceeds a platform switch. Compliance and scalability favor a proven enterprise product.
Where data already exists across systems but no integration surfaces it. A centralized hub pulling from QDirect, Tecsys, and CashBag eliminates manual reconciliation without replacing functional systems.
Where the back-end is stable but the experience creates friction and overreliance on institutional knowledge. The four critical heuristic failures are addressable through redesign alone.
The buy/build/refresh framework kept the recommendation from collapsing into a long backlog. Each tool was evaluated independently against technical debt, vendor viability, compliance posture, and the specific pain profile of the users who depended on it.
Framing this as a strategic question, rather than a UX one, gave leadership a decision structure they could act on with clarity. Each recommendation carried business rationale, user evidence, and architectural feasibility.
The framework also surfaced something important: two of the three tools had different problem types. One needed replacement. One needed redesign. One needed connection. Those distinctions changed the solution entirely.
New users complete requests without calling anyone.
If users need side notes, the system has failed.
Users should know what matters before operations are impacted.
Leadership had a buy/build/refresh decision framework with full supporting evidence, not a list of UX improvements. Each recommendation came with technical feasibility, user impact, and cost-of-inaction analysis.
The research documented the workarounds, undocumented dependencies, and operational knowledge that had lived exclusively in people, making the system's real operating cost visible for the first time.
The integrated hub design connected data that already existed across three systems, eliminating the manual reconciliation step consuming up to 25% of key users' working hours every single day.
What users say and what users need are rarely the same thing. The gap between them is where the real design work lives.
Internal enterprise tools carry a particular risk: users who have adapted to failure will score a broken system highly, because their baseline has shifted. Self-reported satisfaction is not a proxy for usability.
The buy/build/refresh framing also changed how the work was received. A UX recommendation inside a business strategy framework reaches different stakeholders, at different levels of the org, and carries more weight in budget and roadmap conversations.