Golden Data

JD Case Study · Gong

People Analytics Manager

Where
Austin · Chicago · NYC · Salt Lake City · SF
Pays
$115K–$175K

Gong sells an AI operating system that makes revenue teams legible to their leaders. This role is the same product pointed inward: the analytical layer between Gong’s workforce data and its leadership’s people decisions. Workday, CultureAmp and Greenhouse as sources, dbt and Snowflake as the stack, dashboards and workforce-planning models as the products. Not a reporting job: a build-the-layer job.

One kind of number here. The analysis below runs on illustrative figures invented to show method. No Gong data is used or implied; the operating logic is the deliverable.

Aug 12, 2026

What they’re actually buying

Three HR systems, one set of numbers

Strip the title down and the mandate is: turn three HR systems into one set of numbers the leadership team trusts. Governed definitions across Workday, CultureAmp, and Greenhouse so headcount and attrition mean one thing in every deck; analytical products stakeholders reach for unprompted instead of queueing for an analyst; data narratives an executive can retell; and an AI-adoption program that automates the workflow without ever automating away the trust. The hard part is named right in the posting — data quality gaps and data governance — because people data is the most sensitive data a company holds, and adoption dies the first time a number leaks or two decks disagree.

The first 90 days

Reconcile, ship the layer, automate one workflow

The order matters. Nobody should change a system they haven’t reconciled, and nobody should present a verdict before the numbers under it can be trusted.

Days 1–30

Listen, then reconcile the definitions

Sit with the Director of People Analytics, People Ops, and the Systems team; inventory every question leadership actually asked last quarter and which ones died in an analyst queue. Audit the Workday, CultureAmp, and Greenhouse feeds into Snowflake — ownership, refresh cadence, known breaks. Reconcile headcount, attrition, and requisition definitions until one number survives every deck, and map the privacy boundary: which fields may never leave the People team, and what enforces that today.

Days 31–60

Ship the governed layer and the first products

Stand up dbt models over the HR sources with tests attached — not_null, accepted-range, and an allow-list gate on anything that leaves the warehouse, so a definition break or a PII leak blocks the publish instead of shipping it. Rebuild the leadership dashboard set on those governed definitions with lineage from source to tile, and ship one self-serve product for the single most recurring stakeholder ask, so the queue starts shrinking by construction.

Days 61–90

Automate one workflow responsibly, and measure it

Pick one recurring reporting workflow and hand it to automation with QC gates, a freshness check, and a rollback — then measure agent-versus-human on cost, latency, and error rate, published as a scorecard rather than claimed in a deck. Close the quarter by delivering the first people-analytics narrative to the leadership team: three decisions the data argues for, each with a number, a confidence, and an owner.

Signature analysis · illustrative data

Signature analysis: the question-to-decision funnel

The product move: treat people analytics as a product with a conversion funnel, not a queue with a backlog. Every drop-off has a different owner — systems integration, metric governance, product design, narrative craft — so the roadmap the posting asks for falls straight out of the decomposition: fix the biggest drop first, and report the funnel itself to leadership as the function’s own health metric.

How many people questions reach a decision, and where do the rest die?Indexed to the first stage

Workforce questions leadership asks100indexed to 100 — a quarter’s worth
Answerable from the source systems78
Survive definition reconciliation55one headcount number, every deck
Answered self-serve, no analyst queue34
Land as a documented decision22the number that funds the function
Illustrative figures to demonstrate the method. No Gong or employee-confidential data is used; the operating logic is the deliverable.

Requirement → proof

The posting, answered line by line

Each line of the posting against something already built and running on this site. Where the fit is a ramp rather than a match, the row says so.

What Gong asks forWhat I’ve already shipped
dbt, Snowflake — and robust data governance over sensitive people dataA governed dbt layer running live on this site: anonymizing models, tested definitions, and a leak gate that blocks publishing when a PII pattern or a broken derivation shows up.
Analytical products stakeholders trust and actively use — dashboards, self-serve reportingA self-serve salary explorer over a hand-verified openings dataset — a live workforce-data product, not a mock.
Workforce planning modelsA capacity model that separates staffed, ramped, and productive headcount, then traces each gap into dollars — workforce planning with the math shown.
Data narratives for senior leadership — technical and non-technical audiencesWorkforce economics told as narrative: fifty years of the productivity–pay gap, every figure EPI-sourced, written for the room rather than the analyst.
Responsibly adopt AI and ML to scale analytics and automate workflowsAn agent-run delivery fleet with QC gates, schedules, and a freshness dead-man’s-switch — plus a published agent-vs-human scorecard measuring the automation instead of asserting it.
Data privacy and compliance requirementsThe leak gate’s allow-list design: a column is invisible publicly until a model explicitly selects and reduces it, and a red test blocks the publish. Privacy as CI, not policy.
Experience with HR data ecosystems — Workday, CultureAmp, GreenhouseA competitive teardown of UKG Pro People Analytics and the buy-the-layer-not-the-suite case built against it — the domain map this role operates on, published before the interview.

The same engine

Every requirement above runs on one engine — governed dbt definitions, a leak gate on sensitive data, self-serve analytical products, agent-run automation with QC gates, and narratives built for the room. I have already shipped it in public over the most sensitive dataset I have: my own. Point it at Workday, CultureAmp, and Greenhouse instead and the machine is the same — the funnel gets a workforce question at the top, the leak gate guards the people data, and leadership gets one number that means one thing.

Read the People Analytics teardown →

The hard part of this role isn’t the dbt or the Snowflake — it’s that people data is the one dataset where a single leak or one pair of disagreeing decks ends adoption for a year. That’s a governance-first build, and this page is the free sample of how I’d run it. — Paul Brown