Portfolio Case Study · Google APAC Recruiting

Designing a Candidate Experience System at Scale

How a four-person team rebuilt the candidate experience program for Google APAC Recruiting — from candidate-facing prep, to a recruiter enablement microsite, to an anonymity-preserving analytics dashboard, to a quarterly recognition system that kept the work alive.

Role
End-to-end contributor
Team
4 people, cross-functional
Scope
APAC Recruiting org
Disciplines
Service · UX · Analytics · Program
1First touch 2Prep 3Interviews 4Decision 5Close
TL;DR

Problem

Google measures recruiters on a Candidate Experience OKR derived from anonymous post-process surveys. Recruiters scored poorly but couldn't diagnose where in the journey they were losing points. The score was a verdict without a path of action.

Insight

Failures were distributed across the entire candidate journey — confusing first-touch comms, weak prep, broken closing rituals, panel variability, silent feedback loops. The recruiter was the face of all of it but owned only a fraction. Fixing the score meant fixing the system.

What we built

A four-pillar system anchored by an operational ritual: (1) candidate-facing prep on day one, (2) a recruiter microsite as single source of truth, (3) a cascading anonymity-preserving analytics dashboard, (4) a quarterly recognition program tied to the OKR — all reinforced by (5) weekly pipeline hygiene that mechanically forced the "art of closing" behavior.

Why it worked

Each piece reinforced the others. Candidates arrived less confused → recruiters had cleaner conversations → the dashboard surfaced the remaining gaps → recognition rewarded the behaviors that closed them. A closed loop, not a one-off intervention.

01 · Context

Candidate experience is a first-class metric at Google.

Every candidate — offered or not — receives a post-process feedback survey. Responses feed an OKR that recruiters and recruiting leaders are evaluated against. The bar matters most for non-offered candidates, because their experience is the purest test of how the team behaved under pressure.

At APAC scale, that craft is genuinely hard. A single recruiter may run dozens of pipelines simultaneously. Closing well — managing timing, delivering a "no" with care, protecting a candidate's dignity — is a discipline that quietly degrades as load increases.

The failure surface was scattered — that was the real problem.

Through Ops data and conversations with recruiters and candidates, the failure points showed up everywhere, not anywhere.

Auto-rejection emails
Sent in cases that warranted a personal call.
Weak prep material
Arriving late, incomplete, or mismatched to the role.
Decisions in limbo
Results sitting too long while candidates guess.
Missed calls
Candidates don't pick up, then later feel ghosted.
Feedback expectations
Candidates want detail that policy doesn't always permit.
Panel variability
Outside recruiter control, inside candidate memory of "Google".
Zero process visibility
Candidates didn't know GCA, RRK, or how many rounds were coming.
Emotional rejection
Even well-delivered feedback isn't always accepted.

Two characters. Two paths. One score.

Meet the candidate and the recruiter at the heart of every CE score — then watch what happens when the system works versus when it doesn't.

Aanya
L4 Software Engineer · Bengaluru · 5 yrs experience
Wants
A fair, transparent shot at Google — and to know where she stands at every step.
Fears
Wasting weeks of prep only to vanish into a silent inbox.
Success
A respectful close — yes or no — that lets her plan her next move.
Touchpoints
5 emails · 4 interview rounds · 1 closing call
Ravi
APAC Eng Recruiter · Singapore · 4 yrs at Google
Wants
To run a clean, kind pipeline — and hit his CE OKR without burning out.
Fears
A low CE score he can't diagnose. A candidate he forgot to close.
Constraints
40+ active pipelines. Panel availability. Policy on feedback he can share.
OKRs
CE score · turnaround ratio · stale pipeline cleanup · closing quality

The happy path vs. the frustration path.

Same five stages. Same candidate. Two completely different scores at the end.

1First touch
2Prep
3Interviews
4Decision
5Close
Happy path · system working
Welcoming intro email
Process map + prep doc arrive day one.
Confident prep
Watches the videos, knows what GCA is.
Engaging panels
Calibrated interviewers, on time, on topic.
Proactive update
Ravi messages: "Decision by Friday."
Personal call
A real conversation — yes or no, but heard.
Frustration path · system failing
?
Bare-bones email
No process map. No round names. Just a date.
Guessing in the dark
Googles "what is GCA" the night before.
Inconsistent panels
One rushed, one late, one asking the wrong thing.
Silence
14 days. No update. No reply to her email.
NO
Auto-rejection email
A generic template. No context. No call.

Candidate sentiment, stage by stage.

What candidates actually said in post-process surveys — clustered by stage. The faces are illustrative; the patterns came from real qualitative analysis with the Ops team.

"Finally — someone explained the process upfront."
Positive
sentiment +0.78
"Knew exactly what each round would test. Felt prepared."
Confident
sentiment +0.65
"Mixed — one interviewer was great, one felt rushed."
Mixed
sentiment +0.12
"It's been two weeks. I emailed twice. No reply."
Frustrated
sentiment −0.71
"An auto-email. After four rounds. That's how I found out."
Hurt
sentiment −0.84
Pattern: sentiment turns at stage 4 (decision) and is locked in by stage 5 (close). That's the lever.
The reframe

The recruiter is the face of the entire experience but owns only a fraction of its failure points. That asymmetry is the real problem — and it's a system design problem, not a recruiter performance problem.

02 · Research

We talked to both sides — recruiters and candidates.

Two research tracks in parallel. Recruiters could describe the symptoms with precision. Candidates could describe the experience but rarely had the language for the cause.

Who we listened to

Candidates we spoke with spanned roles, regions, and stages of the funnel.

APAC
Region focus
4+
Functions covered
70/30
Non-offer focus
Where the signal is purest
L3–L7
Seniority mix
1:1
Interview format
Strongest signal

A large share of candidate frustration showed up before the first interview even happened. Candidates had no map of the process — no idea how many rounds, what each round tested, how long they'd wait, or what "GCA" meant. That uncertainty primed them to read every later silence as neglect.

It feels like getting a weather report every quarter. I know the temperature, but no one tells me if I need to bring an umbrella or a coat.
Recruiter, APAC · on receiving their CE OKR score · paraphrased
I didn't even know there were four rounds. I would've prepared completely differently if someone had just told me at the start.
Candidate, post-process · non-offered, Engineering pipeline · paraphrased

Recruiter interviews

Recruiters didn't dispute their scores. They couldn't act on them. The "weather report" framing came up again and again — knowing the temperature without knowing what to do about it.

Candidate conversations

Confusion at the front of the funnel poisoned sentiment at the back of it. By the time the candidate received a decision, the experience verdict was already written.

Ops data deep-dive

Working with Recruiting Ops, we mapped every survey question to a stage, ran correlation against overall score, and identified the anonymity boundaries we couldn't cross.

03 · The system we designed

Four pillars plus a weekly ritual. One closed loop.

Each piece was useful on its own. Together they formed a system where measurement, education, enablement, recognition, and operational hygiene reinforced one another.

Candidate Experience PILLAR 1 Candidate Prep Day-one orientation PILLAR 2 Microsite Hub Single source of truth PILLAR 3 Cascading Dashboard Anonymity-preserving PILLAR 4 Recognition Program Sustains the lift a closed loop, not a one-off intervention

Before, the parts existed. They just didn't talk to each other.

The candidate experience program existed on paper. In practice, every piece of it lived in a different doc, a different team, a different cadence.

Before

A scattered set of well-meaning parts.

  • Templates lived across 12+ Google Docs, with no canonical version
  • Prep material was inconsistent — some recruiters wrote their own, others sent nothing
  • CE scores arrived quarterly, with no breakdown by stage or pattern
  • Stale pipeline cleanup happened ad hoc, when someone remembered
  • Recognition was informal — a Slack message here, a shoutout there
  • Leaders saw aggregate numbers but couldn't act on them
  • Recruiters knew their score but not their gap
After

A connected system with one source of truth.

  • One curated microsite, versioned templates, function-tailored
  • Standard prep doc sent on day one — same shape for every candidate in a function
  • Stage-level dashboards refreshed weekly with diagnostic prompts
  • Friday pipeline cleanup as a non-negotiable ritual across all teams
  • Quarterly recognition with explicit, behavioral criteria and public celebration
  • Leaders, managers, and recruiters each see the right slice for their role
  • Every score now points at a specific behavior to change

Five principles we designed against.

Before building any of the pillars, we wrote down the rules the system had to obey. Every later decision was tested against these.

01

Anonymity is sacred.

Candidate trust depends on honest feedback. No view, drill-down, or recognition mechanic can ever expose who said what.

02

Visibility scales with responsibility.

Each role sees exactly enough to act and no more. Leaders get patterns, managers get team signal, recruiters get personal diagnostics.

03

Diagnose, don't blame.

Every score must come with a "so what do I change?" path. A verdict without a diagnostic is corrosive.

04

Reward the behavior, not the outcome.

Recognition celebrates sustained craft — closing well, turning around fast, showing up for the candidate — not lucky scores.

05

Close the loop, every time.

Measurement only matters if it feeds back into education, enablement, and recognition. A one-way dashboard is a dead dashboard.

PILLAR 01

Candidate-facing prep, sent on day one.

The fix for front-of-funnel confusion was upstream, not downstream. We built function-tailored prep material and sent it in the very first email after a candidate entered the pipeline.

  • Separate versions for Engineering, Product Manager, Program Manager and other functions
  • Explained the standard interview architecture: Googleness & Leadership, GCA, two RRK rounds
  • Realistic timelines between stages — so silence wasn't read as neglect
  • Embedded YouTube prep links to Google's own public content
  • Validated with real candidates before rollout. Feedback was strongly positive.
📧 New message
G
Your interview with Google
Recruiting · today
Welcome — here's everything you need.
A quick map of what to expect, what each round tests, and how to prepare.
ROUND 1
Googleness & Leadership
ROUND 2
GCA
ROUND 3
RRK · Part 1
ROUND 4
RRK · Part 2
Watch: prep video for your function
recruiting-hub.googleinternal/apac
HomeTemplatesRecognitionAnalytics
One-stop-shop for recruiters & leaders

Templates, announcements, recognition, and analytics in one place.

v 2.3 · updated today
Template · Eng
Outreach & prep email
Template · PM
Closing-the-loop script
Template · PgM
Rejection with care
Announcement
New CE OKR cadence
⭐ Wall of Recognition
Q3 honorees announced
Process
Standard interview timeline
PILLAR 02

A recruiter & leader microsite — one source of truth.

Before the microsite, templates, guidelines, and announcements lived in dozens of places. Recruiters spent real time hunting for the right version of the right artifact.

I designed and built the microsite end-to-end. It treated internal enablement as a product, not a wiki — because wikis decay and products get curated.

  • Repurposable templates per function — outreach, prep, closing, rejection
  • Process timelines and the candidate journey map (so recruiters and candidates read from the same script)
  • Announcements and program updates
  • The Wall of Recognition — Q-over-Q honorees
  • Direct links into the analytics dashboard

Microsite information architecture

Designed so any artifact is no more than 2 clicks from the home page.

⎈ Recruiting Hub · APAC
Home — landing, role-aware quicklinks static
Templates
By function — Eng · PM · PgM · Sales · UX static
By stage — Outreach · Prep · Scheduling · Closing · Rejection static
Tone library — formal · warm · brief static
Process & Timelines
Candidate journey map static
Round-by-round expectations static
SLA reference (per stage) static
Announcements — program updates, CE OKR cadence dynamic
Wall of Recognition
Current quarter honorees live
Past honorees archive dynamic
Nomination form static
Analytics → deep-link into the CE dashboard external
Help — feedback, change log, who-owns-what static
PILLAR 03

A cascading dashboard — diagnosis without exposure.

This solved the original "verdict without diagnostic" problem. Role-aware views cascaded responsibility down the org while preserving candidate anonymity at every level.

The cascade is the design idea. A leader who notices a dip asks their managers. Managers see which stage and sub-team is dragging the average and have a coaching conversation. Recruiters see their own diagnostic and know what to focus on next week. Organizational gravity moves the metric — not exposure.

Deliberately rejected

  • Individual recruiter leaderboards — collapses psychological safety
  • Raw candidate comments to recruiters — could identify candidates in low-volume pipelines
  • Real-time score-drop alerts — punitive tone, panic cadence
L
Leader view
APAC-wide score · sub-region, function, team breakdowns · trends · hotspots. No individual recruiter visible.
M
Manager view
Team aggregate · broken down by stage (scheduling, prep, interview, decision, close). Anomaly flags, no individual callouts.
R
Recruiter view
Own aggregate over time, by stage. Diagnostic prompts: "Your prep-stage score is below team average — here's what candidates flag most often at this stage."
ce-dashboard.googleinternal · Leader view
APAC CE Score
82
▲ 7 vs Q1
Above target
74%
▲ 18%
Response rate
61%
▲ 4%
Score by funnel stage · APAC
100 75 50 25 First touch Prep Schedule Interview Decision Close
This quarter Last quarter

User flow: how a CE issue actually gets fixed

The cascade in motion — from anonymous candidate signal to a recruiter changing their Monday behavior.

CANDIDATE Submits survey response OPS LAYER Anonymizes + tags by stage DASHBOARD Aggregates by team & stage LEADER VIEW Spots APAC dip in close stage MANAGER VIEW Sees Team 3 trail aggregate dropping RECRUITER VIEW Sees own close-stage prompt + tip OUTCOME Monday behavior changes candidate signal → anonymized → cascaded → coaching conversations → behavior change
PILLAR 04

A quarterly recognition program — sustains the lift.

Visibility surfaces problems. Recognition sustains the fix. We designed this program from scratch to keep candidate experience alive across quarters.

  • A Google Form went to leaders each quarter to nominate team members
  • Criteria emphasized sustained behavior — three months of consistency across:
    • Turnaround ratio
    • Art of closing the loop
    • Proactive responsiveness to candidate queries
  • Winners celebrated publicly — events, newsletter, Wall of Recognition on the microsite
  • Tied directly to the CE OKR — material, not symbolic

The mechanism we cared about was social, not financial. Public recognition of specific behaviors teaches the whole org what "good" looks like.

⭐ Wall of Recognition · Q3
QUARTERLY HONOREES
P
Priya · Eng Recruiting
Closing the loop · 3-quarter streak
A
Arjun · PM Recruiting
Turnaround ratio · top decile
M
Mei · PgM Recruiting
Candidate responsiveness · 100%
Nominated by leaders · validated against dashboard metrics
PILLAR 05 · OPERATIONAL RITUAL

Weekly pipeline hygiene — the forcing function behind "art of closing".

You cannot close well on a candidate you've forgotten about. Stale pipelines were one of the quiet killers of candidate experience: a candidate sitting in "reviewing" for six weeks is a candidate already deciding Google doesn't care.

We made weekly pipeline cleanup a non-negotiable ritual across every recruiting team. Each Friday, every recruiter walked their pipeline and made a decision on every stale candidate — move them forward, schedule a call, or close the application with a thoughtful note.

Why this mattered to CE

  • Cleaning the pipeline mechanically triggered closure of dormant applications
  • Closure forces communication — which is exactly where "art of closing the loop" lives
  • Communication, even on a "no", is the single biggest lever on non-offered candidate sentiment
  • Cleanup discipline was itself one of the OKRs — so the ritual rewarded the right behavior twice

The dashboard surfaced stale pipelines as a leading indicator. Recruiters could see their own queue, managers could see team-level staleness, and the Friday cleanup became a shared cadence rather than an individual chore.

Monday · before cleanup
5 stale
A
Aanya · SWE L4
42 days
B
Bhavna · PM
38 days
K
Kenji · PgM
35 days
M
Mei · SWE L5
29 days
P
Priya · UX
12 days
Friday · after cleanup
0 stale
A
Aanya · SWE L4
closed · called
B
Bhavna · PM
moved to offer
K
Kenji · PgM
closed · called
M
Mei · SWE L5
scheduled
P
Priya · UX
12 days
04 · Process & iterations

What we tried before we landed here.

The system above looks tidy. The path to it wasn't. Here are the bigger explorations we tested and rejected — and what each one taught us.

rejected RECRUITER LEADERBOARD 1. R. Sharma 2. M. Tan 3. K. Lee

Individual recruiter leaderboard

An early concept: rank every recruiter by their CE score publicly. Theory: visibility = motivation.

Killed · Broke principle 01 (anonymity could be reverse-engineered in small teams), collapsed psychological safety, and incentivized gaming over genuine craft.
rejected ! Score dropped 12 pts · last 7 days REVIEW NOW

Real-time score-drop alerts

Email or Slack alerts to recruiters and managers the moment a score dipped. Theory: faster feedback = faster fix.

Killed · Punitive tone, panic cadence, no time to reflect. Replaced with a weekly review rhythm — calm enough to learn, frequent enough to matter.
rejected RAW CANDIDATE COMMENTS identifies candidate

Surface raw candidate quotes

Show recruiters verbatim candidate comments. Theory: nothing teaches like the candidate's own words.

Killed · Even sanitized, free text can identify a candidate in a small pipeline. Replaced with stage-level diagnostic prompts written by Ops.
rejected handwritten per candidate × 200

Bespoke prep email per candidate

Have recruiters write a tailored prep note for every single candidate. Theory: maximum personalization = best experience.

Killed · Doesn't scale to 40 pipelines per recruiter. Replaced with one excellent function-tailored template + room for a short personal line.
rejected FLAT DASHBOARD · ALL USERS 76 82 71

One flat dashboard for everyone

A single shared view: same data for leaders, managers, recruiters. Theory: transparency through sameness.

Killed · Broke principle 02. Leaders had too much detail, recruiters too little context. Replaced with role-aware cascading views.
shipped CASCADING + DIAGNOSTIC Leader · 82 ▲ Manager · 76 ▼ Recruiter · prep stage prompt

What shipped: role-aware cascade

Each role gets exactly the slice they need — enough to act, never enough to identify a candidate or single out a recruiter.

Shipped · Honored all five principles. Created the top-down nudge loop that actually moved behavior on Mondays.

Design decision log

A snapshot of the key trade-offs across the system. Every "chose" was tested against the five principles.

Decision Chose Rejected Why
Score visibility Role-aware cascade Flat shared dashboard Different roles act on different signals. Sameness ≠ transparency.
Surfacing candidate voice Stage-level diagnostic prompts Raw candidate quotes Anonymity is sacred. Prompts give coaching value without identifiability.
Feedback cadence Weekly review rhythm Real-time score alerts Reflection cadence beats panic cadence. Calm enough to learn.
Recognition mechanism Quarterly behavioral nominations Public leaderboard Reward sustained craft, not lucky scores. Behavior is teachable; outcomes aren't.
Prep material Function-tailored standard doc Bespoke per-candidate notes Scalable consistency > unscalable personalization at 40+ pipelines per recruiter.
Microsite model Curated product, versioned Open wiki, anyone-edit Wikis decay. Products get loved. A single owner per page kept quality high.
Pipeline hygiene Friday weekly ritual, team-wide Individual quarterly cleanup Shared cadence creates social accountability. Quarterly is too slow for a candidate sitting in limbo.
What this section is really saying

Every design choice was a real trade-off, tested against the five principles. The system you see in the previous section is the version that survived. The graveyard above is bigger than the result — which is exactly how it should be.

05 · UI craft

The pixel layer.

Strategy and structure only matter if they land in something usable. Here's the design system, the component library, and the screens those decisions became.

Design tokens

The foundation: type, color, and spacing scales the whole system was built on.

Typography
Display / 40 Sustained lift
H1 / 28 Section heading
H2 / 20 Sub heading
Body / 15 The recruiter is the face of the experience.
Caption / 12 Supporting label
Mono / 11 apac.score · q3
Color palette & usage
Primary actions, links, leader signal
#1A73E8
Critical risk, manager signal, problems
#EA4335
Caution recruiter signal, attention
#FBBC04
Success outcomes, recognition
#34A853
Ink primary text, chrome
#202124
Surface panels, neutral fills
#F1F3F4
Spacing scale · 4-pt base
4 · xs
8 · sm
12 · md
16 · base
24 · lg
32 · xl
48 · 2xl
64 · 3xl

Component library

A small, intentional set of components used across the microsite, dashboard, and recognition surfaces. Every screen in this case study composes from these.

Buttons · 3 styles
Used for primary CTAs (e.g., "Open dashboard"), secondary nav, and text-only links inside cards.
Status pill
On track At risk Off track Live
Surfaces stage health on the dashboard and stale-status on pipeline rows.
KPI tile
APAC CE Score
▲ Lift
The atom of the dashboard. Label, value, optional delta. Used in 12+ contexts.
Avatar · gradient identity
AP
RV
MK
+5
Gradient seeded from initials. Used on Wall of Recognition, pipeline rows, persona cards.
Stage badge
1
First touch
3
Interviews
5
Close
Anchors the candidate journey throughout. Color = stage health, number = position.
Input · text
Default and focus states. Used on the microsite search and the nomination form.
Surface card
Template · Outreach
Base container for everything from templates to dashboard panels.
Toast · feedback
Template saved.
Subtle, persistent ~3s. Confirms actions without blocking the user.

States: every screen has four lives

The recruiter view of the dashboard in its four states. Detail design is what separates "looks like a mock" from "feels like a product."

Prep stage
▲ Lift
vs. team avg
DefaultHealthy data, expected case.
LoadingSkeleton, not spinner. Preserves layout.
No data yet
Once you've closed 5 candidates, your prep-stage score appears here.
EmptySets expectations + next action.
Couldn't load data
Ops sync delayed. Refresh or check status page.
ErrorExplains, recovers, never blames.

Hero shot: leader view of the dashboard

Composed entirely from the components and tokens above. Real fidelity — at the level it shipped.

CE DashboardAPAC Recruiting
Leader view
Teams
Stages
Recognition
Pipeline hygiene
Settings
APAC · Q3 review

Candidate Experience · Leader view

APAC CE score
▲ Lift
▲ vs Q2
Above target
↑ More
▲ vs Q2
Response rate
↑ Higher
▲ vs Q2
Stale pipelines
↓ Fewer
▼ vs Q2

Score by funnel stage

Last 4 weeks ▾
high low target First touch Prep Schedule Interview Decision Close
This quarter Last quarter

Teams to watch

Stage drop ▾
TeamStageStatus
Eng · Bangalore Close Off track
PM · Singapore Decision At risk
PgM · Tokyo Prep At risk
UX · Sydney On track
Sales · Mumbai On track

Hero shot: the recruiter microsite

One-stop-shop for templates, recognition, and announcements. Designed as a product, not a wiki.

Recruiting Hub · APAC
RV
Everything you need to close well, in one place.
Function-tailored templates, recognition, and the metrics that matter — curated for APAC recruiters.
v 2.3 · updated today
⭐ Featured · Q3 Recognition
Honorees announced
Three recruiters recognized for sustained closing-the-loop excellence across Q3. See the wall →
Template · Eng
Outreach & prep email
Function-tailored, candidate-validated.
Template · PM
Closing-the-loop script
For the personal call, not the auto-email.
Process
Standard interview timeline
Round-by-round expectations and SLAs.
Announcement
New CE OKR cadence
Weekly review · monthly summary.
Analytics
Open the dashboard →
Your view of stage scores and prompts.

Designed across breakpoints

Same recruiter view, three viewports. Hierarchy and information density adapt; meaning doesn't.

1440 × 900 · desktop
My CE view
SCORE
RANK
Top
STALE
2
Desktop · 1440Full sidebar, multi-column KPIs, in-line charts.
768 · tablet
My CE view
SCORE
▲ Lift
STALE
2
Tablet · 768Sidebar collapses to top tabs, KPIs stack 2×2.
375
My CE view
SCORE
▲ Lift
PROMPT
Prep stage below team avg
Mobile · 375Single-column. KPI → prompt only.
Why this section exists

The thinking is in the rest of the case study. The craft is here. A working system is the sum of both — strategy you can defend and a surface you can ship.

06 · ROI

How this program created value.

Candidate experience work is almost always undersold financially because the biggest returns are diffuse — brand, referrals, re-applications, recruiter retention. The case gets stronger when the system closes the loop.

CE score trajectory · pattern observed

Shape of the trend over the rollout window · exact figures held under Google confidentiality
Sustained lift
across quarters
higher lower target Prep launch Microsite live Dashboard rollout Recognition Q1

Direct returns

Measurable, attributable, and visible inside the org.

Hours reclaimed weekly per recruiter

Productivity reclaimed

The microsite consolidated artifacts that previously lived across dozens of docs and inboxes — pulling search and reformatting time out of the recruiter's week.

Sustained OKR lift across quarters

CE score movement

A directional, persistent improvement in candidate experience scores across APAC over the rollout window. Specific figures held under Google confidentiality.

"where are we?" emails

Less candidate silence

Front-loading clarity reduced mid-funnel anxiety emails — work that drained recruiter time and tanked sentiment.

Indirect returns

Harder to attribute, but where the real long-term value sits.

Employer brand protection

A non-offered candidate is a future user, customer, referrer, and re-application. Improving their experience protects the brand at a scale no marketing budget can match.

Referrals & re-applications

Candidates who leave a process feeling respected reapply and refer. Candidates who leave feeling ghosted do neither.

Leadership signal quality

CE went from a number in a slide to a discussion in QBRs. Leaders could connect process changes to score movement and justify CE investment.

Recruiter retention & morale

Being measured on a metric you can't diagnose is corrosive. A clear path of action plus public recognition changed how the role felt to do.

07 · The personal arc

This is the project where I found UX.

I came into this work solving a recruiting operations problem. I left it knowing I wanted to design for a living.

The moment it clicked

I was on the fourth iteration of the leader view of the dashboard. It was late on a Wednesday. I'd been pushing the team aggregate card around the canvas for an hour, trying to figure out why it still didn't feel right — and I realised the problem wasn't visual at all. It was that the card was answering the wrong question. A leader doesn't open a dashboard to see a number; they open it to know where to look. So I rewrote the whole hierarchy: hotspots first, score second, trend third.

That night I went home and noticed two things. First, I'd forgotten to eat dinner. Second, that I'd been looking forward to doing exactly that kind of work every morning for months — the small, infuriating, generative loop of "this doesn't feel right yet, let me try again." The microsites I had been building before were the seed. That Wednesday was when I stopped calling design a side interest and started calling it my craft.

My role, over time

How my contribution evolved across the lifecycle of the program.

Phase 1 · Discovery
Operations & analytics partner

Started by partnering with Recruiting Ops to make sense of the survey data — mapping questions to stages, finding the correlation patterns, understanding the anonymity rules.

Phase 2 · Design
Information designer & microsite builder

Took on the microsite end-to-end — IA, page design, content strategy, curation model. Began co-designing the cascading dashboard with Ops.

Phase 3 · Program
Co-architect of the recognition program

Helped design the nomination criteria, the cadence, the public-recognition mechanics. Brought in candidate validation for the prep material.

Phase 4 · Stewardship
Loop-closer across all four pillars

The role that emerged: connecting analytics signal back to program decisions, keeping the system coherent as it scaled across APAC.

Constraints we designed around

Anonymity was the sharpest, but it wasn't the only one.

Anonymity

Every view, drill-down, and recognition mechanic tested against the rule: candidate identity must never be back-solvable.

APAC scale & locale

Multiple regions, multiple languages, multiple work cultures. Templates and timelines had to flex without breaking the shared standard.

Policy on feedback

What recruiters can and cannot share with non-offered candidates is bounded by policy. The system had to support kind closure within those bounds.

Stakeholder consensus

Recruiting Ops, Legal, Recruiting Leadership, and individual hiring teams all had a vote. Every design moved through that consensus.

What I'd do differently

Honest reflection

I would have pushed harder on panel accountability.

The candidate experience score is shaped enormously by the interview panel — interviewer quality, calibration, on-time starts, tone in the room. We treated the panel as outside the recruiter's control and therefore outside the system we built. In hindsight, that was the wrong call.

A meaningful chunk of candidate sentiment was being driven by an inconsistent panel experience, and we built no surface for it. Done again, I would have proposed a parallel feedback loop for panels — anonymous, cascading, with its own diagnostic prompts — and made the case that "candidate experience" isn't really one thing the recruiter owns; it's a shared craft between the recruiter and the panel.

The system we shipped lifted the recruiter side of the score. The panel side stayed essentially untouched. That's the unfinished business.

What I personally owned

We were a team of four. I want to be honest about scope:

Skills this work drew on

Service design UX research Information architecture Data visualization Behavior design Content design Analytics & measurement Program design Stakeholder management Microsite design & build Workshop facilitation System thinking

About me

RB
Rashmi B S
UI/UX Designer at Iwanttfc · designer crossing over from candidate experience & ops
A note on what's not shown. Out of respect for Google's confidentiality, this case study does not include the actual microsite or dashboard screens, real OKR scores, the verbatim candidate questionnaire, or any data that could identify a candidate, recruiter, or internal team. All visuals here are illustrative reconstructions of the structure and design rationale — which are mine to share. The proprietary surface is not. Happy to walk through any of the thinking in more depth in conversation.