Stage dwell times
Precise time-in-stage per candidate derived from the event stream, not from a manually updated status.
Product engineering
The Walt
Where candidates move, and where they stall.
A recruitment analytics platform that turns an applicant tracking system's raw event log into a picture of where a hiring funnel actually leaks, with the bottleneck named rather than left to be inferred.
- Sector
- HR Tech & Recruitment
- Year
- 2024
- Duration
- 5 months
- Team
- 5 people
- Platforms
- Web · Email integrations
- Status
- In production
246
Candidates in pipeline
Across 18 open roles
21d
Average time-to-hire
Down from 34 days at rollout
11d
Longest stage stall found
Screen to onsite scheduling
18
Open roles tracked
Each with its own baseline
01 — The problem
The talent team ran a competent ATS and still could not answer basic questions. Time-to-hire was reported as a single company-wide average that hid a three-week gap between engineering and commercial roles.
Worse, nobody could see stalls while they were happening. A candidate sitting untouched for eleven days between a technical screen and an onsite only became visible when they withdrew.
02 — What we did
We rebuilt the funnel as a state machine over the ATS event stream, so every candidate has a precise dwell time in every stage rather than a status field that was last updated whenever someone remembered.
Then we made the product argue. Instead of rendering a chart and leaving interpretation to the reader, The Walt names the stage costing the most days, quantifies it against the team's own baseline, and links to the specific candidates sitting in it right now.
A dashboard that does not name the bottleneck is just a prettier spreadsheet.
Interface
What it looks like in use
Two of the screens that carry the most weight in daily use, rebuilt here from the production design system.
Hiring funnel · EngineeringBOTTLENECK
Applied412
Screen · 2.1d avg186
Screen to onsite · 11.4d avg74
Onsite · 3.0d avg31
Offer12
Scheduling costs 11 of 21 days
The product names the bottleneckRather than rendering a chart and leaving interpretation to the reader, the stage costing the most days is stated in words.
Needs action today18 ROLES
Stalled 12d · Staff EngineerHIGH
Stalled 9d · Platform LeadHIGH
Awaiting panel · Data Engineer4d
Offer pending · SRE2d
246IN PIPELINE
21dTIME TO HIRE
Ordered by riskThe daily view lists exactly which candidates need action today, ranked by how close they are to withdrawing.
Capabilities
What the system does
Grouped by the job each set of capabilities exists to do, rather than by which team built it.
Funnel analytics
01 · The core measurementConversion by stage
Pass-through rates per stage, per role and per source, with confidence bands so small samples are not over-read.
Cohort comparison
This quarter against last, and this role family against the company baseline.
Source attribution
Which channels produce candidates who actually convert, rather than which produce the most applications.
Bottleneck detection
02 · The product's opinionNamed bottlenecks
The stage costing the most days is stated in words, with the size of the cost and the candidates affected.
Stall alerts
Candidates exceeding the dwell threshold for their stage are surfaced while it is still recoverable.
Interviewer load
Scheduling delay traced to specific panel availability instead of blamed on the process in general.
Drop-off diagnosis
Withdrawals correlated with the stage and dwell time that preceded them.
Team workflow
03 · Getting it acted onRecruiter dashboard
A per-recruiter view of exactly which candidates need action today, ordered by risk.
Hiring manager digest
A weekly email that states what moved, what stalled and what needs a decision.
Role scorecards
Structured evaluation captured against consistent criteria so comparisons are meaningful.
ATS write-back
Actions taken in The Walt sync back so the ATS stays the system of record.
Architecture
How it is put together
Layer by layer, with the reason each one exists — because the reason is usually the interesting part.
01Ingestion
Connectors poll ATS webhooks and history endpoints, reconstructing a complete event stream including backfilled history.
- Webhooks
- Backfill jobs
02Funnel engine
A state machine that replays events into per-candidate stage transitions, producing dwell times that survive out-of-order delivery.
- Event sourcing
- PostgreSQL
03Analytics
Pre-aggregated rollups per role, stage and cohort, refreshed incrementally so dashboards stay fast as history grows.
- Materialised views
- dbt
04Insight layer
Rule-based detection of stalls and bottlenecks, scored against each team's own historical baseline rather than an industry average.
- Baseline scoring
- Alerting
05Application
React dashboard with server-driven charts and a digest mailer.
- React
- Visx
- Postmark
Technology
Frontend
- React
- TypeScript
- Visx
- Tailwind
Backend
- Node.js
- NestJS
- PostgreSQL
- Redis
Data
- dbt
- Materialised views
- Event sourcing
Platform
- AWS
- Docker
- GitHub Actions
- Datadog
Outcome
What changed
Measured against how the operation ran before, not against a benchmark chosen after the fact.
-
Time-to-hire fell from 34 days to 21. Most of the gain came from one scheduling stall the team had never been able to see.
-
Averages stopped hiding role differences. Each role family is measured against its own baseline.
-
Stalls became recoverable. Alerts fire while a candidate is still engaged rather than after they withdraw.
-
The ATS stayed the system of record. Write-back means the team did not end up maintaining two truths.
How it ran
-
Weeks 1-2
Data audit
Assessed ATS event completeness and found the history gaps that had to be backfilled.
-
Weeks 3-6
Funnel engine
State machine, event replay and dwell-time computation validated against manual reconstruction.
-
Weeks 7-12
Analytics and insights
Rollups, baselines and the bottleneck detection rules.
-
Weeks 13-18
Workflow surfaces
Recruiter dashboard, digest mailer, scorecards and ATS write-back.
-
Weeks 19-22
Rollout
Phased by department with baselines established before any target was set.
Related work
From the same practice
Recycling & E-Waste
Razor ERP
One connected system for a warehouse that ran on paper.
A mobile-first ERP for warehouse and e-waste operations that replaced a decade of spreadsheets, paper travellers and radio calls with a single connected system that works on the floor..
2024 · 7 months
Investment & Asset Management
Kilwa
Market data, sentiment and research in one investment workspace.
An AI investment-intelligence platform that pulls market data, news sentiment and analyst research into a single workspace, so an investment team can go from a question to a defensible answer without leaving the screen..
2025 · 9 months
Healthcare & Clinical Communication
Lyfe Languages
Clinical translation where a wrong word has consequences.
A translation platform built specifically for clinical settings, where general-purpose translation is not safe enough and a verified medical dictionary sits between the model and the patient..
2023 · 8 months
Something similar?
Running into the same problem The Walt solved?
Tell us what you run. We will reply within two business days with what we would build for your situation — and, just as usefully, what we would leave out.