•
The data roadmap - discovering what product, marketing, ops, and finance need from data, prioritizing it against platform health, and sequencing the investment. You’ll present it, defend it, and re-plan it as the business moves.
•
Data quality and freshness - automated monitoring across source data, pipelines, and reports; catching upstream schema and source changes before they break anything downstream; running incidents to resolution when they happen.
•
Data lineage and impact analysis - a living map from production source to warehouse model to dashboard, and the process that uses it: when a production change is proposed, its downstream impact on pipelines, metrics, and reports gets assessed before it ships, not discovered after. The end-state is data contracts with engineering, so breaking changes get caught in their workflow, not ours.
•
Lightdash - administration, workspace structure, permissions, and the rollout itself. Your job is to give the company self-serve autonomy while keeping the workspace tidy enough that people can find and trust what’s there. Enablement is part of the deal, people follow standards they’ve been taught, and so is keeping queries fast and warehouse costs sane.
•
The semantic layer - we just shipped it for our most critical metrics: one governed definition per metric, in code. You’ll extend definition and mapping to the rest and guard the layer against uncontrolled growth as it scales.
•
Event tracking governance - our governed Segment event catalog: reviewing new events against its standards, keeping it matched to what production actually sends, and evolving the guardrails (naming, property dictionary, drift detection) as tracking grows.
•
AI data readiness - AI agents query our warehouse every day through Brain, our internal AI toolkit. You’ll govern what data AI tools can access and keep the warehouse AI-legible: documented, consistent, and safe for an agent to query and get the right answer.
•
Data security and privacy - access controls, PII handling and retention under US state privacy laws, and periodic reviews of who, and which AI tools, can see what.
•
The governance system itself - the documentation, ownership models, and review loops that keep all of the above running without heroics.
Turn business needs into a data roadmap Every area of the company wants something from data, and today those asks reach the Analytics team as a stream of interruptions. You’ll build the intake and prioritization that turns them into a roadmap , one that balances stakeholder needs against platform health, survives contact with a changing business, and that your stakeholders can see themselves in. The hard part: saying “not yet” to important people, with a reason they respect.
Make the Lightdash migration a step-change, not a re-platforming We’re replacing Tableau and Metabase with Lightdash. Done poorly, we trade two messy tools for one messy tool. You’ll design the structure, spaces, permissions, certification, naming, that lets stakeholders self-serve at the speed the company needs without creating an uncontrolled dashboard-growth nightmare. The hard part: autonomy and tidiness pull in opposite directions, and you have to deliver both.
Finish and defend the semantic layer We just shipped our semantic layer for our most critical metrics, one governed definition per metric, so “two dashboards, two numbers” can’t happen. The unglamorous truth: a long tail of metrics still needs definition and mapping, and a semantic layer only stays trustworthy if someone curbs its growth. You’ll own both, extending coverage and keeping one-metric-one-definition true as the layer scales.
Tame event-tracking entropy Segment events power our funnels and product analytics, and they’re implemented by many engineers across many teams. The guardrails exist, a governed event catalog with naming standards, a property dictionary, a review lifecycle, and automated drift detection against production. What’s missing is a dedicated owner: someone who holds every new event to the standard, keeps the catalog matched to what production actually sends, and evolves the guardrails as tracking grows. Without that, entropy wins, events drift and silently degrade when features change.
Get ahead of breakage instead of chasing it Today, when production data changes upstream, we too often find out when a pipeline breaks or a stakeholder flags a wrong number. You won’t start from zero, an AI-powered Analytics Engineer agent already runs freshness monitoring, metric anomaly detection, and dbt-based lineage checks, but it doesn’t yet run at the scale or coverage we need. You’ll take detection from partial to comprehensive, extend lineage beyond dbt (Segment events and Lightdash need stitching in), and wire it into engineering’s change review, so a proposed production change comes with a downstream impact assessment instead of a postmortem. The end-state is data contracts: breaking changes caught in engineering’s workflow, not ours.