Redesigning the tool hundreds of employees use every day.

How rethinking ODIN — OnGrid's internal verification platform — cut task time by 39%, nearly halved errors, and turned the lowest-rated internal tool into the most-improved one.

ODIN redesigned dashboard
Company OnGrid
Role Lead Strategy & Design
Team 1 Designer, 3 Engineers
Duration 6 months

What is ODIN, and who uses it?

ODIN is OnGrid's internal command centre — used exclusively by OnGrid employees to manage the entire lifecycle of background verification requests. It's not a public product. It's the operational backbone hundreds of people rely on every single working day.

Operations Executives

Handle high volumes of requests daily. Need speed, scannability and minimal friction to hit throughput targets.

Account Managers

Manage client-specific views, track SLAs, and coordinate between internal teams and external vendors.

Team Leads

Monitor team performance, triage escalations, and ensure quality across verification workflows.

A tool built for the system, not the people using it.

ODIN had been built incrementally to mirror backend data structures — not employee workflows. Over time it became a sprawling, inconsistent tool that hundreds of employees were forced to adapt to, rather than one that adapted to them.

Old ODIN — the main request table before redesign

What employees were saying:

"

The tool is difficult to navigate. I can never find the specific feature I'm looking for.

"

The design is cluttered and overwhelming. It's hard to focus on the information I need.

"

It often crashes mid-task. It's a real hassle to use.

"

It doesn't feel designed for our team's needs. It feels like it was just thrown together.

Survey results before redesign

40 respondents · 75% response rate · 1–5 scale

Ease of use
1.5
Design & aesthetics
1.8
Speed & performance
1.9
Overall satisfaction
2.0
Customization options
2.0
Support & documentation
2.1
Features & functionality
2.2

Understanding how employees actually worked.

Before touching any design, I spent three weeks in the field — interviewing, observing, and surveying. The most important discoveries didn't come from surveys. They came from watching people work.

Interview session or observation during ODIN research
12

In-depth interviews

Across Operations, Account Management and Team Leads. Semi-structured, focused on daily tasks and workarounds.

8

Observational sessions

Watched employees complete real tasks in real time — surfacing workarounds they had fully normalized.

40

Survey respondents

11 Likert-scale questions across 7 usability dimensions, async via Slack. 75% response rate.

3 root causes behind every pain point

Developer's model vs. user's model

ODIN was built to mirror backend data structures, not user workflows. Every feature added on top of that foundation inherited the same mismatch — and every pain point traced back to it.

Three teams, one undifferentiated experience

Operations Executives needed speed at volume. Account Managers needed client-specific views. Team Leads needed progress oversight. The tool treated all three identically.

Systemic UI inconsistency

Actions, labels, and interaction patterns varied unpredictably across different sections. Users had to re-learn conventions in each part of the tool — doubling cognitive load.

9 changes across the platform.

Each solution maps directly to a research insight. Nothing was redesigned for aesthetic reasons alone — every change was driven by observed behavior, error patterns, or measured friction.

Solution 01

Search & result customization

The problem

Search was one-size-fits-all. Every session began with the same manual filter reconfiguration — before any real work could start. Operations Executives, Account Managers, and Team Leads each needed completely different result views, yet the tool served them identically. A repeating tax on every user, every day, before they could even begin their actual task.

Configurable filters, column preferences, and result sorting are now saved per user role. The system remembers the setup. Users arrive at their working context immediately — filters pre-applied, columns relevant to their workflow, results scoped to what they actually need. The experience shifted from system-driven to user-driven, aligning search behaviour with individual task priorities.

↓ 39% avg. task time Role-specific defaults
Before
Default search — one size for all
After
Customized search panel
Solution 02

EDU document viewer

The problem

Education verification required verifiers to juggle two completely separate contexts. The candidate's data — institution name, dates, degree title — lived on one screen. The supporting documents — scanned certificates, transcripts, mark sheets — lived on a different portal entirely. Verifiers had to open the verification details, navigate away to the document portal, memorise what they saw, return to ODIN, and manually compare from memory. This context-switching wasn't just slow — it increased verification errors. Working from memory while comparing details is a reliable way to miss discrepancies. On top of this, uploaded documents frequently arrived as rotated scans or poorly cropped images. Without any built-in document tools, verifiers downloaded files, opened them externally, and worked entirely outside ODIN to correct orientation before they could even begin verifying.

The redesigned Education Verification workspace consolidates everything into a single view. Candidate-entered data and uploaded supporting documents are displayed side by side — verifiers see the degree certificate and the institution name field simultaneously, without switching screens. Cross-referencing is now a visual task, not a memory task. Document utility tools — rotation, zoom, and crop — are integrated directly into the viewer. Verifiers correct orientation and focus on the relevant portion of a document without leaving ODIN, without downloading files, and without breaking their verification flow. Related actions are grouped in the same workspace, so the entire task — review, compare, verify — happens in one place.

Zero context switching Docs + data in one view Built-in crop & rotate
Before
Separate document portal — verification details in ODIN, documents elsewhere
After
Unified EDU verification workspace with inline document viewer
Solution 03

CCRV — Criminal Case Review & Verification

The problem

A high-stakes, irreversible confirm/reject decision was shown inside the full cluttered interface. Navigation, filters, and background list items competed for attention at the worst possible moment. Users had also normalized using Ctrl+F — the browser's native find tool — to search within the criminal case list. A broken workaround no one had flagged until we watched them work.

A dedicated modal dialog now isolates the decision from all surrounding UI. The background disappears. The user's full attention is on the case details and the confirm/reject action — nothing else. For the search problem, a built-in contextual search tool replaces Ctrl+F, with semantic tokens (case type, date, jurisdiction) so users can search the way they think, not the way the system stores data.

↓ 59% CCRV error rate 8.2% → 3.4%
Before
Old CCRV — full interface visible during confirmation
After
CCRV modal dialog with focused case review
Solution 04

Data sufficiency — from 11 clicks to 4

The problem

Verifying whether a candidate had provided required documents meant opening the request, navigating to a separate candidate portal, reviewing their uploads, then returning to ODIN to mark it sufficient or insufficient. 11 clicks across two portals — for a task performed dozens of times per day.

A contextual dialog inside ODIN now consolidates candidate information, document uploads, and the mark-as-sufficient action in one place. Users review and act without leaving the tool. The journey went from 11 clicks to 4 — and the cognitive overhead of context-switching went with it.

↓ 64% click reduction 11 → 4 clicks
Before
Candidate portal — separate tab from ODIN
After
Data sufficiency dialog in ODIN
Solution 05

Next-action prediction

The problem

Every possible action for a request was displayed in each table row — maximum visibility, but extreme visual noise. Hard to scan, easy to misclick, and consuming disproportionate screen space. With hundreds of requests on screen, the cumulative cognitive load was significant.

Only the 2 most contextually relevant actions are now shown per row — based on the current workflow state. All additional actions are available via a secondary side drawer. Progressive disclosure reduces clutter without hiding flexibility. Less to read, fewer errors, faster decisions on every row in the table.

Before
Table row — all actions visible at once
After
Two primary actions plus side drawer for remaining options
Solution 06

Persistent sticky sidebar navigation

The problem

The hamburger menu hid navigation by default. Every time a user needed to switch sections — a constant requirement in an operational tool — they had to open the menu first. A small tax per interaction that compounded across hundreds of daily navigations into a meaningful productivity drain.

Navigation was converted to a persistent sidebar with visible icons — always accessible, single-click switching, zero extra interactions. The icon-first approach follows a progressive familiarity model: new users rely on visual recognition while exploring; returning users build icon-to-action muscle memory over time. Both ends of the experience spectrum are served without compromising either.

Side menu
Persistent sticky sidebar with always-visible icon navigation

Supporting changes

Solution 07

Color-coded status chips

The problem

Request states were displayed as plain text labels with no visual differentiation. In a workflow where operations staff handle hundreds of requests simultaneously, manually reading and interpreting every status label was slow and mentally exhausting. The lack of visual hierarchy also made it harder to identify what action was needed next — and which requests were blocked, pending, or complete.

Each request state was assigned a distinct colour treatment — a chip that communicates status at a glance without reading. Users can now scan the full request table and instantly understand the state of every row. Cognitive effort dropped, operational awareness improved, and decision-making on which request to act on next became significantly faster — especially during high-volume periods.

Status Chips
Color-coded status chips in the request table
Solution 08

Non-intrusive toast notifications

The problem

Workflow status updates were communicated through frequent modal pop-ups triggered by almost every action — request creation, assignment, status updates, verifier responses. While informative, these interruptions demanded acknowledgment before work could continue. For an ops team performing dozens of actions per hour, the cumulative disruption to task flow was significant.

Status updates now appear as non-intrusive toast notifications anchored at the bottom-left of the interface — visible, but never blocking. They appear and fade without requiring any user action. The system remains transparent about what's happening without pulling users out of their current task. Awareness is preserved; interruption is eliminated.

Notifications
Bottom-left toast notification
Solution 09

Surfaced high-frequency toolbar

The problem

Three actions — Add Note, Data Suggestion, and Bug Reporting — were used repeatedly throughout every operational session. All three were buried inside secondary menus, requiring users to navigate away from their current task every time. Research revealed these weren't occasional utilities — they were core parts of the daily workflow, used multiple times per hour by every ops team member.

All three high-frequency actions were surfaced as a persistent contextual toolbar within the primary workspace — always visible, always one click away. Keeping essential tools at the surface of the interface eliminates the navigation overhead that had been quietly draining productivity across every session. Users stay in flow; the tool comes to them.

Toolbar
Persistent contextual toolbar with high-frequency actions

Measured 3 months post-launch.

Baseline metrics were established before the redesign launched. The same instruments — session timing, system logs, observation, and the re-run survey — were used to track change over the following quarter.

39%
Task completion time

9.2 min → 5.6 min across core workflows

59%
CCRV error rate

8.2% → 3.4% — the highest-stakes action in the tool

64%
Clicks per data sufficiency check

11 clicks → 4 clicks

42%
Support tickets / month

38 → 22 navigation and confusion tickets

40%
New employee time-to-productivity

3 weeks → 1.8 weeks for independent operation

95%
Overall satisfaction score

2.0 → 3.9 out of 5 — same survey, same respondents

173%
Ease of use score

1.5 → 4.1 out of 5 — the lowest baseline score, highest improvement

~320 hrs
Saved per month

Estimated across high-frequency workflows across the operations team

What this project taught me.

Six things I carry into every project since.

01
Observation reveals what surveys can't

The Ctrl+F workaround in CCRV — one of the biggest wins — would never have surfaced in a survey. Users had normalized it. It took watching them work in real time to notice it was a workaround at all.

02
Architecture is upstream of everything

Every pain point traced back to information architecture built around backend data, not user goals. Getting this right early — before adding features — is the highest-leverage design decision you can make.

03
Small friction compounds

The stickiest changes — the sidebar, color chips, toast notifications — were individually modest. No single one transformed the product. But together they removed friction from interactions repeated hundreds of times per shift.

04
Choose metrics before design, not after

I chose success metrics at the start, tied directly to project objectives. This made the impact story credible — not post-hoc justification.

05
Research creates buy-in before design review

Sharing satisfaction scores and user quotes with engineers and the PM before presenting solutions created shared problem ownership. When I proposed changes, the problem was already understood.

06
Design for the whole user journey — novice to expert

The icon sidebar needed to work for a first-week hire and a two-year veteran simultaneously. The progressive familiarity model served both groups without compromising either.

Want to see more?

Get in touch.