Senior Data Scientist — LTV, Bidding & Fraud Modeling

Scrambly S.r.l.; Polska; Remote; PLN 180,000 - 280,000; Full time

Job summary

Scrambly is seeking a Senior Data Scientist to own modelling for how the platform spends and earns money, forecasting user value and keeping forecasts accurate as the product evolves.

You will lead end-to-end modelling, collaborate with engineers on data pipelines and serving, and work with marketing, monetisation, and risk teams in a small, autonomous data team reporting to the Head of Data.

Qualifications

• 4+ years as a data scientist working on production models that drive real decisions.
• Hands-on experience with LTV or ROAS prediction — forecasting value from early behaviour.
• Adtech or mobile performance marketing background with attribution and MMP data.
• Strong Python and SQL with production experience on a cloud data warehouse (BigQuery preferred).
• Rigour about evaluation and out-of-sample discipline.
• Experience of models that made it into production and were used to make real decisions.
• Commercial judgement and clear communication.
• A proactive, self-directed way of working.
• English at B2+, written and spoken.

Responsibilities

• Value prediction: forecast user and cohort value from early behaviour and improve accuracy.
• Bidding and pricing: model spends to acquire users within margin limits.
• Segmentation and targeting: build scoring to treat users and inventory by value.
• Fraud and abuse: improve fraud model accuracy and avoid false positives.
• Evaluation: define model success metrics and monitor performance after launch.
• Collaboration: work with Head of Data, engineers, and commercial teams to convert model output into decisions.

Skills

LTV/ROAS prediction

Python

SQL

Production models

Tools

BigQuery

Job description

About The Company

About The Company

Scrambly is one of the fastest-growing adtech startups in the world, featured in the AppsFlyer Performance Index and Singular ROI Index in just 3.5 years.

We’re growing 250%+ YoY across revenue, team, product, and technology — fully bootstrapped and profitable.

We’re building the future of app discovery: a reward-powered alternative to the App Store and Google Play. Our loyalty-driven ecosystem connects millions of users with the world’s top apps and delivers unmatched, ROI-focused growth for mobile advertisers.

Our mission is to build a true alternative to traditional app stores — fueled by rewards and data — and create a new, performance-first growth engine for mobile app advertisers.

About The Role

We are looking for a Senior Data Scientist to own the models behind how Scrambly spends and makes money — what our users are worth, what we should pay to acquire and reward them, and how we keep the ecosystem free of abuse.

The centre of the role is predicting user and cohort value — forecasting what people will be worth from early signal, and keeping those forecasts accurate as the product and the market move underneath them. Around it sits a set of connected problems across marketing, monetisation, and risk, each of which turns a prediction into a decision the business acts on. We will go through the specifics with you during the process.

You will own the modelling end to end: framing the problem, building the model, deciding how it is evaluated, and staying with it once it is live. You will not be doing that alone — you will work closely with our engineers on data, pipelines, serving, and integration.

This is an early hire on a data team being built from scratch, reporting to and working day to day with the Head of Data. We are flexible on level. The title reflects the ownership and independence the role carries, not a hard experience bar — we will set the level, title, and compensation to match what you bring.

Key Responsibilities

• Value prediction. Own and improve how we forecast user and cohort value from early behaviour, across our different revenue streams. Improve accuracy where it matters most commercially, and keep the models working as the product changes.
• Bidding and pricing. Build the modelling that decides what we are willing to pay to acquire users, together with the controls that keep spend inside margin limits while the system is still learning.
• Segmentation and targeting. Build the scoring and segmentation that lets us treat users, partners, and inventory according to the value they actually deliver.
• Fraud and abuse. Improve the accuracy of our fraud models, and balance catching abuse against wrongly penalising real users.
• Evaluation. Decide how each model is judged and hold that line — out-of-sample discipline, backtesting, and choosing metrics that reflect the commercial objective rather than the convenient one. Stay close to your models once they are live, and flag when accuracy starts to slip.
• Collaboration. Work with the Head of Data on prioritisation, with our engineers on data pipelines and serving, and with commercial teams on turning model output into decisions they trust.
  
  

Requirements

• 4+ years as a data scientist working on production models that drive real decisions. We care more about what you have owned than about the number.
• Hands-on experience with LTV or ROAS prediction — forecasting user or cohort value from early behaviour, and dealing with the problems that come with it: immature cohorts, long horizons, sparse segments, and models that decay as the product changes. This is the core of the role and we are looking for someone who has done it before.
• Adtech or mobile performance marketing background — attribution and MMP data, cohort and campaign economics, and how UA spend actually gets decided. Backgrounds in gaming, gambling, or user-incentive and loyalty systems are equally relevant, and experience with rewarded, offerwall, or incentivised apps is a significant advantage.
• Strong Python and SQL, with production experience on a cloud data warehouse — BigQuery preferred.
• Rigour about evaluation: you choose the metric that matches the business objective, you are careful about out-of-sample and point-in-time correctness, and you are willing to say a model is not good enough to ship.
• Experience of models that made it into production and were used to make real decisions — you understand what it takes to get there and what happens to a model afterwards.
• Commercial judgement and clear communication — able to explain a model to people who will spend money based on it, and to push back when the data does not support the question being asked.
• A proactive, self-directed way of working. You will be in a very small team, with a lot of latitude and little process.
• English at B2+, written and spoken — you will document your work and work daily with an international team.
  
  

Nice to Have

• Bidding, pricing, or budget allocation systems — control loops, bandits, or other approaches to spending under uncertainty.
• Fraud, anomaly detection, or trust & safety modelling, especially with weak or incomplete labels.
• Experimentation and causal inference — A/B design, power analysis, and incrementality measurement.
• Survival, hierarchical, or Bayesian methods applied to retention, monetisation, or sparse-segment problems.
• BQML, Dataform, or dbt, or other experience building models close to the warehouse.
• Deploying and monitoring models yourself — pipelines, serving, retraining, and drift monitoring.
  
  

What We Offer

• Ownership of the models that set UA spend, offer payouts, and fraud decisions — work directly attached to revenue, with visible impact within weeks.
• Problems with real depth: long-horizon forecasting, adaptivity, pricing under uncertainty, and fraud with imperfect ground truth.
• A close working relationship with the Head of Data, a real say in how the data function is built, and room to grow as the team expands.
• A level and compensation matched to what you bring, rather than to the title on the advert.
• A profitable, bootstrapped company growing 250%+ YoY, where investment in data is justified by results rather than budget cycles.
  
  

By submitting this application, I agree that my personal data will be collected, processed, and retained by the company solely for the purposes of managing and assessing my candidacy.