Garrett Gignac · Data & Analytics Leader

I help organizations bring clarity to complex problems

Most of my work starts with an ambiguous question. The toolkit varies: experimentation, causal inference, product analytics, AI. The goal is the same: get to a decision the team can trust.

View selected work How I work

Start with the decision,
not the methodology.

The problems I like most are messy: the answer isn’t obvious, the data isn’t perfect, and another dashboard won’t solve it. I figure out what we actually need to learn, build the right measurement around it, and bring technical and business teams together on what the evidence says and what to do next.

Selected work

Decisions, not deliverables.

A few examples of the problems I’ve worked on, how I approached them, and what changed.

01 · UBER

Building causal measurement for B2B CRM

Problem Attribution told us who got credit, not whether the communications actually changed behavior. In B2B, smaller samples, heterogeneous users, and noisy outcomes made the answer even harder to pin down.

Approach I built an experimentation framework around randomized holdouts, then used variance-reduction techniques to get more reliable estimates from noisy B2B data.

Impact I brought Data Science, Marketing, and senior leaders together around the framework, giving the teams a shared, credible way to evaluate CRM performance and decide what to scale, change, or stop.

ExperimentationCausal inferenceCUPEDB2B measurement

Synthetic monthly effect estimates with confidence intervals

CUPED uses pre-treatment behavior to explain away some of the noise, improving precision without changing the causal question.

02 · TRIPADVISOR PLUS

Connecting supply to consumer product outcomes

Problem More participating hotels didn’t automatically mean a better traveler experience. What mattered was whether a relevant hotel search could actually surface a compelling Plus offer.

Approach I measured how often relevant hotel searches surfaced a Plus offer, then partnered closely with the B2C Product Analytics team to understand what that meant for the traveler experience.

Impact That gave the B2B team a clearer view of where Plus inventory was strong and where coverage gaps could weaken the product experience.

Product analyticsMarketplaceB2B2C

03 · AI-ENABLED PRODUCT

From campaign alert to AI-assisted diagnosis

Problem The alerts worked. The hard part was figuring out why a campaign was underperforming, which still meant piecing together context by hand.

Approach Using AI-assisted development, I built a tool that investigates underperforming campaigns using internal knowledge, campaign context, and performance data.

Impact It created a faster path from ‘something is wrong’ to ‘here’s what to investigate next,’ turning an alert into a tool for diagnosing problems and deciding what to do about them.

AI-assisted developmentAI diagnosisAnalytics products

EXPERIMENTS & IDEAS

How should AI change experimentation?

Problem Teams don’t just run out of traffic. They lose time testing weak ideas and relearning things the company already knew.

Approach My view is to use AI around the experiment: recover past learning, pressure-test hypotheses, catch design issues, and monitor experiment health. Live experiments still determine causal impact.

Potential impact Learn more from each test without lowering the bar for evidence.

ExperimentationAICausal inference

How I think

The loop I come back to.

The methodology serves the question. The analysis serves the decision.

Decision

What does the business need to decide?

Uncertainty

What do we need to learn?

Evidence

What would change our mind?

Design

What is the simplest credible approach?

Action

What changes because of the answer?

Causal measurement

Did X actually cause Y?

A model can’t create causality. The design determines how much confidence we can put in a causal claim.

Pick the situation you’re in.

Expertise

Problems I know how to solve.

Experimentation

A/B tests · holdouts · power · guardrails · heterogeneous effects · incrementality · CUPED · variance reduction

Causal inference

Counterfactuals · quasi-experiments · confounding · difference-in-differences · matching

Product analytics

Funnels · adoption · retention · segmentation · subscriptions · marketplaces

Marketing science

CRM · attribution · paid media · campaign analytics · LTV/CAC

Business analytics

Unit economics · KPIs · forecasting · strategic analysis · investment decisions

AI-enabled analytics

Rapid prototyping · analytical products · workflows · decision support tools

About

Technical rigor. Business judgment.

Most of my career has been spent on questions that don’t have a clean answer: what changed, why it changed, whether we actually caused it, and what to do next.

I’m at my best when the question is messy, the stakes are real, and there are multiple teams in the room. I’ve led analysts and data scientists, partnered across Product, Marketing, and business teams, and translated technical work for senior leaders. I care less about using a particular method than about framing the right question, choosing the right level of rigor, and getting people aligned on the decision.

Start a conversation

Working on a hard data problem?

I’m currently exploring senior roles across Data Science, Analytics, Decision Science, and Product Analytics.

Email me LinkedIn ↗

Examples are generalized from prior work. Any data shown is illustrative or synthetic, and no confidential or proprietary information is included.