UX & Localisation · 2022

Quick Facts

• Industry: Global accounting marketplace (Accounts Reconciliation + Transaction Matching solution)
• Role: Senior Consultant — Marketplace localisation
• Timeline: ~3 months end-to-end, three shipped passes
• Team: Solo lead, with a UK SME panel (three practising controllers and a Big-Four-trained reviewer) validating each pass
• Impact: UK adoption moved from lagging the global baseline to in line with it, region-specific support tickets fell from frequent to rare, and new-user onboarding collapsed from days to hours.

---

Overview

A great global product can be a bad fit locally, until you do the work. The marketplace solution was a mature Accounts Reconciliation and Transaction Matching offering used across multiple regions, shipping with sensible global defaults built around the conventions of its biggest user base. UK-based users adopted it, tolerated it, and quietly built spreadsheets next to it. The brief: close the gap between the platform's defaults and the way a UK finance team actually works — without forking the solution and without waiting for a global release cycle. We shipped a layered localisation: terminology, configuration, workflow, output formats. By the end of the engagement, UK adoption was in line with the global baseline and the support queue had largely stopped hearing the region-specific complaints that triggered the project.

The Problem

The friction wasn't one big thing. It was twenty small ones that compounded.

The global defaults assumed conventions — cost-centre coding patterns, GL hierarchy depths, reconciliation period boundaries, statement format expectations, even the labels on the buttons — that UK practitioners didn't share. None were wrong; they were just foreign. A UK controller opening the platform saw fields named for concepts they'd call something else, templates structured for a chart of accounts that wasn't theirs, and reconciliation windows that didn't line up with how their close was scheduled.

The visible symptoms:

• Adoption was lagging in the UK region. Monthly active users trailed equivalent regions despite comparable rollout effort. Region leads were politely raising it on every quarterly call.
• The support queue had a UK-shaped pattern. Tickets clustered around the same themes: "the default period boundary is wrong for our close," "the matching template asks for fields we don't carry," "this terminology doesn't mean what your tooltip says it does." Most ended with some variation of "this isn't how we do it."
• User confidence in the outputs was suffering. Statement formats didn't look like statements a UK auditor would expect, so reviewers second-guessed the numbers. Status labels used unfamiliar phrasing, so signoff stalled while users asked colleagues what each status meant.
• A shadow process had grown up alongside the platform. UK teams exported outputs into Excel and rebadged them before circulating internally, because the native outputs didn't pass an internal sniff test — defeating half the point of having the platform.

The brief was unfussy: make the UK experience feel native, without forking the product. Three months. Access to configuration, templating, and string overrides — no core code or data-model changes, none needed.

Process

The work split into three phases — overlapping in calendar time, disciplined about what each was trying to learn or ship.

Phase 1 — Enumerate the friction

Three weeks of structured interviews with UK practitioners across the rollout cohort — controllers, AR analysts, reviewers, and one auditor seeing the outputs downstream. Format deliberately narrow: walk me through your last reconciliation, button by button, and flag every time you had to pause, look something up, or translate what the screen said into what you meant. No leading questions; just observation.

The output was a single long list — eventually 60-odd items — of every place the defaults asked the user to do extra cognitive work:

• "The status label Pending Substantiation means nothing to me. We'd say Awaiting Evidence."
• "The default reconciliation period closes on the 25th. Ours closes on the last working day."
• "The matching template wants a Memo field. We carry that in Narrative and there's no equivalent column in our source extract."
• "The auto-generated statement shows accounts in numerical order. UK audit packs lead with the P&L then balance sheet, grouped by control account."

Some items were terminology, some configuration, some workflow steps with no UK analogue, some format expectations downstream of the platform entirely.

Phase 2 — Categorise into four buckets

I sorted every item into one of four buckets, in roughly increasing order of effort to fix:

1. Terminology — string-level overrides. Labels, tooltips, status names, button text.
2. Configuration — default values and templates. Period boundaries, default account groupings, template field sets, threshold values.
3. Workflow — sequencing and step-count changes. Collapsing approval gates the UK process didn't separate, re-ordering steps to match the UK close timing, removing steps that mapped to a global obligation UK teams didn't have.
4. Format outputs — what the platform produced for consumption outside it. Statement layouts, audit pack ordering, export shapes.

Terminology was almost pure win — high impact per hour, low risk. Configuration was bigger but well-contained. Workflow needed SME-panel validation before anything shipped — the wrong "redundant" step removal would break governance. Format outputs needed the auditor's review specifically.

Phase 3 — Ship in that order

Terminology first, because users would feel the win the day it landed; then build trust by shipping the larger changes after. A localisation project that goes silent for three months and drops a big-bang release loses its sponsor; one that ships visible improvements every fortnight keeps the room.

The cadence was a fortnightly drop, each announced in the UK region's standing call with a short before/after walkthrough. By the time the workflow changes landed in week eight, region leads had seen four shipped improvements and were arguing for the project rather than questioning it.

Solution

The localisation layer that came out of the three phases has four named components, each maintainable independently and each contained within the marketplace solution's standard extension surfaces.

1. Terminology layer

A string-override file mapping the global vocabulary to UK English plus UK accounting vocabulary, keyed by the platform's internal string IDs so it survives upstream version bumps without merge pain. Covers the obvious cases (Memo → Narrative, Pending Substantiation → Awaiting Evidence, Match Confidence → Match Status) and the less-obvious ones (status verbs, tooltip phrasing, error messages, audit log entries). Principle: if a UK practitioner wouldn't say it, the platform shouldn't either. SME-panel signoff before shipping removed the bikeshedding loop entirely.

2. Configuration profile

A UK-default profile — period boundaries aligned to the working-day close, account groupings reflecting a typical UK chart, threshold values calibrated to what UK teams flag at, template field sets mapped to UK source extracts. Selectable at onboarding; global defaults remain available. Switching is a single setting; the data model doesn't change.

3. Workflow shortcuts

Three standard workflow sequences carried steps that existed for a governance obligation the UK process either didn't have or already discharged elsewhere. The shortcuts collapse those steps where safe — for example, the global flow separated Reviewer Assigns Evidence from Reviewer Approves Match because in some jurisdictions those are separate control responsibilities; in the UK process they're discharged by the same person at the same moment, so the collapsed flow asks once instead of twice. Each shortcut was paired with a written justification reviewed by the SME panel; any step the panel was unsure about was left in place. Collapse only where the UK process is demonstrably equivalent, never probably equivalent.

4. Format outputs

UK-style statement templates plus an audit pack layout that matches what a UK auditor expects. Statements lead with the P&L then balance sheet, grouped by control account; reconciliations export with the narrative column alongside the balance column rather than in a separate block; the audit pack carries the period boundary and approver signature in the header where reviewers look first. The auditor on the panel signed off each template against a real audit pack they'd recently issued — fit for purpose, not just familiar.

All four components ship as configuration and template files inside the marketplace solution's standard extension layer. Nothing in the core was modified; a future global release is picked up by re-running the overlay on top — additive, not invasive.

Results

| Metric | Before (global defaults) | After (UK-localised) | Change | |---|---|---|---| | UK adoption (monthly active users vs comparable regions) | Lagging the global baseline | In line with global baseline | parity restored | | Region-specific support tickets per month | Frequent — top theme on the UK queue | Rare — no longer a top theme | error class largely cleared | | Onboarding time for a new UK user | Days (multiple guided sessions) | Hours (self-serve with the UK profile) | onboarding collapsed | | User satisfaction signal (regional survey) | Low — below cross-region average | High — above cross-region average | reversed |

Soft outcomes:

• The shadow Excel process retired itself. UK teams stopped rebadging exports because the native ones already looked the way they expected — a half-day per cycle nobody had been counting.
• Region leads stopped raising UK fit as a quarterly concern. The agenda item disappeared from their standing call — the kind of soft outcome that doesn't show up on a slide but tells you the project landed.
• A reusable localisation playbook. The four-bucket categorisation and the fortnightly cadence both carried into subsequent regional rollouts and shipped faster as a result.

Learnings

What worked. Categorising friction into four buckets and shipping in increasing-effort order was the structural decision that made the project work. Terminology fixes landed inside the first sprint and bought political room for the slower configuration and workflow work. Each shipped pass made the next easier to scope — users were trusting the project and bringing better-targeted feedback.

What I'd do differently. Recruit the auditor onto the SME panel in week one rather than week four. The format-output pass was the slowest because we discovered halfway through that the audit pack had layout expectations none of the controllers had thought to raise — they were just used to receiving it that way. An auditor in the first round would have surfaced those alongside the workflow points, and the format pass would have shipped a fortnight earlier.

Skill developed. Distinguishing wrong from foreign. The global defaults weren't wrong; they were foreign to a UK user. The temptation in a localisation project is to treat the gaps as defects, which puts the global product owner on the defensive and slows everything. Framing them as fit-for-region adaptations — global defaults preserved, UK overlay on top — kept the conversation collaborative and made the changes easy to land upstream as a configurable profile rather than a fork.