Based in Ottawa, Canada

Fast-growing companies make their most expensive decisions with the least certainty. We change that.

Using causal inference, mathematical optimization, and agentic AI. From your highest-stakes human decisions to the daily micro-decisions that can be automated.

How We Work

Four capabilities. One end-to-end decision stack.

Most AI and data science consultants can describe these techniques. Few can actually build them, implement them in production, and make them work inside a real business. The difference matters - especially when your board is asking hard questions or your operations team is underwater.

We cover the full spectrum - from the $5M capital decision your leadership team makes once a year to the thousands of operational decisions your business makes every day that currently require human attention they shouldn’t need.

01 · Measurement

Causal Inference

Cuts through noise to find what’s actually causing change - not just correlating with it

Platform attribution is self-reported. Correlations are misleading. Before you can make a confident decision, you need to know what’s genuinely driving outcomes in your business - and what just looks like it is. Causal inference gives you that certainty. Designed experiments, uplift models, and treatment effect estimation that tell you the truth about your data.

Used when: you need to know which initiatives, channels, or features are actually generating return before committing more resource to them.
02 · Decision

Mathematical Optimization

Finds the absolute best action to take - accounting for constraints and uncertainty

Once you know the causal drivers, the next question is what to do about it. Mathematical optimization builds the model that finds the best possible allocation, assignment, or configuration given your real constraints - and that holds up even when the inputs carry uncertainty. Not a recommendation. The true optimal answer, with rigorous logic applied.

Used when: you have a high-stakes allocation or assignment decision - budget, inventory, pricing, capacity - that deserves a rigorous answer.
03 · Automation

Agentic AI Engineering

Builds production-ready systems that make data-driven decisions autonomously - so your people don’t have to

Many decisions in your business don’t need a human - they need a reliable, data-driven system that runs consistently at scale. Agentic AI systems handle repetitive decision workflows autonomously: routing, triaging, classifying, responding, monitoring. Built production-ready, tested, and secure - not a prototype that breaks in week two. Frees your team for the work that actually requires judgment.

Used when: your team is making the same types of decisions repeatedly at volume and that time would be better spent elsewhere.
00 · Foundation

Data Modelling & Pipelines

The infrastructure that makes everything else possible - clean, reliable, production-grade

Causal inference, optimization, and agentic AI engineering are only as good as the data beneath them. Before any of the above can work, the data needs to be accessible, trustworthy, and structured correctly. We build the pipelines, data models, and integration layers that turn raw, fragmented data into a reliable foundation for every decision your business makes.

Used when: your data is siloed, inconsistent, or simply not in a shape that supports rigorous analysis or automation.
10 years
Data science experience
30%
Operational savings through optimization
50+
Models built across marketing, ops and analytics

The gap between knowing and deciding is where margin disappears.

Most fast-growing companies have data. Many have invested in measuring what’s driving performance. The problem isn’t the measurement - it’s that the measurement doesn’t drive the decision.

And below the big strategic decisions, there’s a second problem: hundreds of daily operational decisions that consume skilled people’s time, get made inconsistently, and could be handled by a well-built system.

Both problems can be fixed. That’s what we do.

Your attribution is self-reported

Meta and Google both claim credit for the same conversions. You have no independent view of what’s actually driving revenue - and budget decisions get made on numbers that are fundamentally compromised.

Measurement doesn’t drive the actual decision

Companies invest in incrementality tests and then allocate next quarter’s budget in a spreadsheet by instinct. The bridge between the measurement and the decision has never been built.

Operational complexity outgrows the infrastructure

The warehouse layout, carrier mix, and inventory positioning that worked at $20M becomes a cost ceiling at $80M. Nobody inside has the modeling capability to optimize their way out of it.

Your team is making decisions that shouldn’t need humans

Routing tickets. Classifying requests. Monitoring thresholds. Triggering actions. These decisions happen dozens or hundreds of times daily and consume time your best people should be spending on harder problems.

Work

Proof across the decision stack. Built, not theorised.

Marketing Causal Inference Mathematical Optimization
True lift

From broken attribution to a defensible budget allocation

A fast-scaling brand was allocating a significant marketing budget based entirely on platform-reported attribution. Each channel claimed credit for the same conversions. Leadership suspected the numbers were wrong but had no way to prove it, quantify the waste, or redirect spend with confidence.

We designed and ran a causal measurement framework that isolated the true incremental impact of each channel. Those estimates - with honest uncertainty bounds - fed directly into a robust optimization model producing a quarterly budget allocation the CFO could scrutinise and the CMO could act on. The model reruns each planning cycle as the channel mix evolves.

How it was built:

01 - Attribution audit: identified where platform self-reporting was inflating channel contribution and creating systematic blind spots
02 - Causal measurement design: geo-holdout and intent-to-treat framework calibrated to the specific channel mix and data quality
03 - Estimates with uncertainty bounds: not confident point estimates, but honest ranges reflecting what the data could and couldn’t tell us
04 - Robust optimization model: an allocation performing best even when estimates are imperfect; rerunnable every quarter
“For the first time we had a budget allocation we could actually defend. Not one we’d reverse-engineered from gut feel.”
OperationsMathematical Optimization
30% savings

Warehouse costs reduced through slotting optimization

A scaling distribution operation had outgrown its warehouse layout. Pick rates were declining as order complexity grew. We built a slotting optimization model from scratch - aligning product positioning with pick frequency, order co-occurrence, and labour cost dynamics. Identified 30% in potential savings. Executable recommendation the COO could act on immediately.

Customer AnalyticsCausal Inference
360° view

Proactive retention through lifecycle modelling

The commercial team was reacting to churn rather than anticipating it. We built an account status model giving precise visibility into where every customer was in their lifecycle - identifying early churn signals, expansion windows, and re-engagement moments invisible in the existing reporting stack. Enabled proactive intervention at scale.

AI Engineering
Automation

Agentic AI Systems for Operational Decision Workflows

The pricing team was making the same markdown and promotion decisions every week. Manually, reactively, at a cost in time that didn’t match the complexity of the decision. We built an agentic system that monitors sell-through in real time, evaluates promotional incrementality against a live holdout, and triggers pricing actions autonomously.

About

Built inside real scaling platforms, production grade and reliable.

Lerilo is a boutique data science practice specializing in optimization science applications for the North American market.

The founder spent ten years as Senior Data Scientist at companies ranging from high growth e-commerce to multi-national video entertainment platforms.

Not advising. The person in the data, writing the code, building the models, presenting findings to the COO and GM on decisions that mattered. DC location selection. Warehouse slotting. Carrier network design. Marketing attribution. Customer lifecycle modelling. Actually implementing real high impact solutions that most consultants can only talk about.

Our approach is business first. We bridge the gap between business and deep technical work, presenting results in a way a business audience can understand.

Business first
Rigorous
Actionable
Proven results
Data visualization and analytics
Services

Four ways to engage. One logical place to start.

01 · Start here

Decision Audit

Fixed cost $2,000

A structured review of one decision problem you are currently making by instinct - and what it would take to make it analytically. Works for any function: marketing, operations, retention, pricing, or an automation opportunity.

Deliverables:
  • Review of existing data and measurement infrastructure
  • Gap analysis - what’s reliable, what isn’t, what’s missing
  • Prioritised roadmap for the next step
  • 30-minute findings presentation with Q&A
02 · Core engagement · Human decisions

Analytical Project

From $5,000 / project

A defined analytical project with a clear deliverable. The methodology depends on the problem - causal inference, mathematical optimization, statistical analysis, data integration, or a combination. You receive something working and repeatable.

Deliverables:
  • The appropriate solution implemented for your specific problem
  • A working, tested deliverable - model, system, analysis, or integration
  • Documentation and handover so your team can use and maintain it
  • Findings presentation to relevant stakeholders
03 · Core engagement · Automated decisions

Agentic AI System

From $5,000 / project

A production-ready agentic AI system that handles a defined class of operational decisions autonomously. Built, tested, and secured to production standard. Not a prototype - a system your business can rely on.

Deliverables:
  • Decision workflow mapping and automation scoping
  • Agentic AI system design and build
  • Integration with your existing data and tooling
  • Production deployment, testing, and documentation
04 · Ongoing

Retainer or Maintenance

$3,000 – $6,000 / month

Models need refreshing as conditions change. Agentic systems need monitoring and improving. Ongoing retainer keeps your analytical infrastructure current - and extends the relationship to new problems as they emerge.

Deliverables:
  • Quarterly model refresh with updated data
  • Agentic system monitoring and improvement
  • New problem evaluation as the relationship develops
  • Monthly check-in call
Get In Touch

The first conversation costs nothing and clarifies everything.

Tell me about the decision you’re trying to make better - or the workflow you want to automate. If there’s a clear fit I’ll say so directly. If there isn’t, I’ll tell you that too.

Contact Information

Phone

+1 437 216 2411

Office

Ottawa, Ontario
Canada

Base & Hours

Ottawa, Ontario - working with fast-growing companies across Canada and the US

Mon – Fri, 9am – 6pm Eastern