Data & ML Specialist: Attribution & Revenue Intelligence

About BRILU

BRILU is an AI-powered growth platform focused to support ecommerce brands, B2C servicess brands and marketing agencies.

The role in one line:

You support BRILU's attribution system: the models and pipelines that turn messy, self-

serving platform numbers into one trustworthy answer - which marketing channels actually

drive a client's revenue, and where to move the budget. You're the first and only data hire;

frontend, backend, and devops are already in place around you.

What you'll do

The system has one job: tell a client, with numbers they trust, where their revenue really comes from. You get there in four steps - each one a part you own.

1. Reconstruct customer journeys - so the full path to a sale is visible, not just the slice each

platform shows.

- Build the identity-resolution logic that stitches anonymous sessions into one journey

per user - deterministic with a login, probabilistic without.

- Model event data in the warehouse: sessionize, order touchpoints, produce complete

conversion paths.

- Define the channel taxonomy and the per-vertical event schema.

2. Deduplicate conversions across platforms - so the conversion count is real and matches

the client's actual sales, instead of the inflated sum where Meta, Google, and email each claim

the same purchase.

- Build the matching engine that ties each platform's conversion claim to the one real

conversion (order ID - click ID - identity + time window).

- Collapse duplicate claims into a single canonical record; handle view-through

separately.

- Produce the reconciliation numbers: real conversions vs. the inflated platform sum.

3. Build the attribution engine - so the client learns which channels truly drive revenue and

how to reallocate budget.

- Implement the attribution models: Time-Decay, Position-Based / W-Shaped, Linear,

Markov Chain, Shapley.

- Build the automatic model selector - industry first (picks the model family), then

volume (picks how advanced the method can be).

- Build automatic industry classification as the selector's first step.

- Calibrate the models against incrementality tests (holdout / geo experiments).

- Produce the revenue-attribution output and the data behind budget-reallocation

recommendations.

4. Manage the reports & dashboards

- Deliver the data and logic behind the funnel, SPRTM (proprietary BRILU method), and

revenue-attribution dashboards.

- Report uncertainty honestly - consistent, decision-grade attribution, never a fake 100%

precision.

Requirements

Must have:

- 5+ years in data science / analytics / data engineering, with real modeling work.

- Strong Python (pandas, numpy, scikit-learn) and advanced SQL (window functions,

data modeling).

- Marketing data (Google Ads, Meta, GA4) and business metrics (CAC, LTV).

- Understanding of multi-touch attribution and the difference between correlation and

causation. You don't need to have built an attribution engine, but you must understand

why summing platform numbers is wrong.

- Attribution models (Markov, Shapley) or libraries like Channel Attribution.

- Applied statistics: A/B testing, significance, incrementality.

- Ability to explain an attribution result to a non-technical client whose numbers don't

match Meta Ads Manager.

Nice to have:

- Identity resolution / entity matching.

- dbt, Airflow, BigQuery/ClickHouse, MLflow.

Profile: Leaning toward modeling and ML, not just pipelines. We have the infrastructure; the

differentiator is attribution, and that's where we want depth.

Tech stack: Python (pandas, numpy, scikit-learn, statsmodels) · SQL · multi-touch attribution

(Markov, Shapley, ChannelAttribution) · A/B testing and incrementality · GA4 and revenue

attribution. Nice-to-have: dbt, Airflow, BigQuery/ClickHouse, MLflow.