Lifecycle & growth analytics partner

Know where your retention and marketing spend actually leak — and what is truly working.

I’m Souvik, an independent data science partner for subscription, e-commerce and marketplace teams. I diagnose churn and CRM leakage, prove what’s incremental, and build the ML your team can act on — so every decision moves a number the business cares about.

Book a free 30-min call Take the Lifecycle Leak Scorecard

6+ years in applied ML & growth analyticsEd-tech, travel, marketplace and fintech · working with EU & US teams

30%

lower marketing cost from a lead-scoring model

20%

lead-quality uplift from propensity modelling

15%

revenue improvement from funnel and channel analytics

Millions

of users covered by lifecycle churn and CRM models

Results from prior in-house roles; details anonymised in Selected work.

Is this you?

Five signs you need a thought partner, not another dashboard

Most growth teams don’t lack data. They lack a trusted answer to “what should we do next, and how sure are we?”

CRM and paid reports show strong ROAS, yet total revenue isn’t moving as much as the reports suggest.

You can’t say how many customers a campaign caused to convert, versus customers who would have converted anyway.

Churn is discussed after it happens. Nobody can name the early behaviours that predict it.

Experiments are run, but results get debated instead of used, and nobody trusts the sample size.

Every metric question waits days for an analyst, so decisions get made on gut feel.

Best fit

B2C subscription, e-commerce or marketplace companies, roughly $5M–$100M revenue
A growth, CRM or lifecycle team without a senior in-house data scientist
Buyer: Head of Growth, CRM/Lifecycle, CMO or COO
Data in a warehouse (Databricks, BigQuery, Snowflake) and a CRM tool such as Braze, Klaviyo or Customer.io

Probably not a fit

One-off dashboard or ticket-style tasks with no decision attached
Teams that need a full-time hire or a large delivery squad
Projects where no one can act on the findings

How we can work together

Four ways in, each tied to a business outcome

Fixed scope and fixed fee for audits and sprints; a monthly retainer for ongoing partnership. All scoped on a free call.

Start here2 weeks · fixed fee

Lifecycle Leak Audit

A fast diagnostic of where retention and CRM revenue leak, with a ranked, sized list of what to fix first.

Churn drivers and early-warning behaviours
Segmentation beyond basic RFM
CRM revenue attribution sanity check
Prioritised opportunity map with estimated impact

You get: a readout for leadership, the supporting analysis, and a 90-day roadmap.

2–6 weeks · fixed fee

Measurement & Experimentation Sprint

Prove what’s incremental and set up the system so the team keeps proving it.

Holdout and incrementality test design
Sample-size, duration and decision rules
Bayesian and frequentist readouts your team can interpret
MMM-lite for teams with thin data

You get: a working experiment pipeline and clear go/no-go rules.

Most popular for ongoing workMonthly retainer

Fractional Data Science Partner

A senior data scientist embedded with your growth team for a fixed number of hours each month.

Monthly metrics and experiment review with leadership
ML roadmap: churn, uplift, LTV and propensity models
Async Q&A and decision support between meetings
Mentoring for your analysts, with documented handoffs

You get: one accountable owner for “what do the numbers say, and what next?”

Add-on or standalone

Self-Serve Analytics with AI Agents

Let marketing and CRM stakeholders answer their own metric questions on governed data, without waiting on an analyst.

LLM agent connected to your warehouse through MCP
Domain-knowledge skill: table hierarchy, correct SQL patterns, root-cause playbooks
Guardrails, evaluation and monitoring

You get: faster answers and fewer analyst interruptions. I’ve built and run this approach in production.

The ML toolkit behind the offers

I use the simplest method that moves the metric, and bring ML when it earns its keep.

Churn & propensity (XGBoost, LightGBM) Uplift modelling LTV forecasting Behavioural segmentation Demand forecasting Incrementality & MMM-lite Agentic analytics (MCP)

Selected work

Problem, approach, outcome

From my in-house roles. Company names are withheld here out of respect for confidentiality; happy to discuss on a call.

30%lower marketing cost
20% better lead quality

Global travel marketplace · Lead scoring

Problem: Marketing spent evenly across leads, though most were unlikely to convert.
Approach: Built a propensity model (logistic regression through random forest to XGBoost), handled class imbalance, and tuned the score threshold against cost and revenue rather than accuracy alone.
Outcome: Spend on low-propensity leads fell, cutting marketing cost by 30% while lead quality rose 20%. The team also gained cohort LTV forecasts and weekly executive dashboards for budget decisions.

Millionsof users scored for churn and targeted across email, push and in-app

Global ed-tech platform · Lifecycle & CRM

Problem: Retention campaigns reached everyone alike, and campaign impact was hard to separate from natural behaviour.
Approach: Churn propensity models (XGBoost, LightGBM) on Databricks, K-Means and RFM segments, automated feature pipelines for audience generation, plus holdout and lift experiments read with Bayesian and frequentist methods.
Outcome: Leadership could target retention spend by risk and expected uplift, make go/no-go calls on experiments, and build new audiences far faster. An LLM agent over governed data let stakeholders run their own root-cause analysis.

15%revenue improvement
30% faster reporting

Property marketplace · Growth analytics

Problem: Spend wasn’t tied to which channels and cohorts actually paid back, and the funnel had unexplained drop-offs.
Approach: Channel-level ROAS, cohort LTV and payback analysis, plus funnel segmentation that located the worst drop-off at the “contact agent” step.
Outcome: Spend shifted to top-performing channels and the drop-off led to a UX redesign and an agent response-time SLA. Automated Python reporting cut turnaround by 30%.

Figures are from my own prior roles and not guarantees. Results depend on your data, market and team.

How I work

Decisions first, models second

I start from the decision you need to make and the metric it should move. The method follows.

Frame

A free 30-minute call to agree the decision, the metric, and what “good” looks like.

Diagnose

Read-only access to your data. I find the leaks, size them in revenue terms, and rank them.

Prove

Experiments or models that separate cause from coincidence, with decision rules agreed in advance.

Embed

Documented handoffs, team coaching and a monthly review so gains compound after I step back.

Free tools

Where does your lifecycle program stand?

A two-minute self-assessment, plus a rough estimate of the value at stake. Both run in your browser; nothing is sent unless you choose to share it.

Lifecycle Leak Scorecard

A directional self-assessment, not an audit.

Value-at-stake estimator

If part of your CRM and paid spend isn’t incremental, here’s roughly what that is worth.

Monthly CRM + paid marketing spend (USD) Share that may be non-incremental: 10%

$5,000 / month

about $60,000 a year

Illustrative arithmetic only, not a forecast. The real share is exactly what a measurement sprint would find out.

About

A data scientist who thinks like an operator

I’m Souvik Samanta, a data scientist with 6+ years across ed-tech, travel, real estate and fintech. Today I lead lifecycle analytics and ML for a global ed-tech platform, where I work as a thought partner to CRM and marketing leadership: turning model outputs into targeting strategy, experiment decisions and budget calls.

An MBA-equivalent in finance and analytics gives me the commercial lens. I also founded and operated a fintech product for several years, so I know what it costs to ship, measure and iterate under real constraints.

I consult on the side, about 20 hours a week, which means a small number of clients at a time and my full attention on each. I work on IST, with full overlap on European working hours and early-morning calls for US East.

Modelling & statistics

XGBoostLightGBMScikit-learnProphetBayesian A/BCausal & uplift methods

Data & engineering

PythonSQLDatabricksBigQuerydbtAirflowDocker

CRM & BI

BrazeAmplitudePower BILooker

LLM agentsMCPEvaluation

Education

PGDM (MBA-equivalent), Finance & Analytics, IMI New Delhi

Recommendations

What colleagues say

Verified LinkedIn recommendations from people who worked with me directly.

“I had the pleasure of working with Souvik at Travelopia for over two years. He consistently showcased exceptional skills in ML modeling using Python, delivering impressive results on complex projects. Beyond his technical expertise, Souvik is a remarkable team player, fostering collaboration and encouraging open dialogue among team members. His positive attitude and strong work ethic greatly enhanced our team dynamic. I am confident that Souvik will be a valuable asset to any organization he joins, bringing both expertise and a collaborative spirit to the table.”

Tejash PopateBusiness Analyst, Travelopia · same team · LinkedIn

“I had the pleasure of working with Souvik for more than 1.5 years at Travelopia, and I can confidently attest to his exceptional skills and contributions. Souvik was an invaluable asset to our team, delivering high-quality work on various analytics projects, including building dashboards in Power BI, analysis using Python, and developing machine learning and factor-based models. His dedication, strong work ethic, and technical expertise enabled him to tackle complex projects with ease. What impressed me most about Souvik was his willingness to support and collaborate with other team members, fostering a spirit of teamwork and knowledge sharing. I highly recommend Souvik for any future roles.”

Jyoti KumarAnalytics Leader, Data Science & Predictive Modelling · managed Souvik directly · LinkedIn

Questions

Before we talk

How is this different from hiring a freelancer or an agency?

I’m accountable for the decision, not the ticket. You get a senior partner who frames the question, does the analysis and ML, and helps your leadership act on it. Scope is fixed up front, so you’re buying outcomes rather than hours.

What does it cost?

Audits and sprints are fixed fee; the partnership is a monthly retainer for an agreed number of hours. I scope and quote after a free 30-minute call, once I understand the decision and the data.

What do you need from us to start?

One accountable sponsor, read-only access to your warehouse (or agreed extracts), and a metric you want to move. I can work with Databricks, BigQuery, Snowflake and Postgres, and CRM tools such as Braze.

How do you handle confidentiality and data?

I sign an NDA before seeing anything sensitive. Client data stays in your environment where possible, and I don’t reuse it or share it with other clients.

Time zones and communication?

I’m on IST (UTC+5:30). That overlaps fully with European working hours, and US East morning calls work well for me. West Coast by arrangement. Day to day I work asynchronously with written updates.

Do you work with competitors?

I’ll tell you upfront if I see a conflict of interest, and I won’t share one client’s information with another.

Let’s talk

Have a retention or measurement problem worth solving?

Book a free 30-minute call, or send a short note. I reply within one business day.

Thanks, your message is in. I’ll reply within one business day.

Book a free 30-min call [email protected]Email linkedin.com/in/souviksamanta95LinkedIn github.com/souviksamanta95GitHub

Capacity is about 20 hours a week, so I work with a small number of clients at a time.