Pricing · Capacity · Labor · AI systems

Custom decision systems for businesses whose margins ride on pricing, capacity, and labor.

I build the pricing engines, scheduling optimizers, and forecasts that cruise lines, resorts, and theme parks run on. Larrea & Co. brings that work to mid-size operators and the investors who own them, built for your constraints instead of bought off the shelf.

Larrea & Co. will take on a small number of engagements in the future. Until then I write about pricing, optimization, and applied AI, and I'm glad to compare notes.

Stay in touch What I build

Every cell is a unit of capacity: a cabin, a room, a ticket, a shift. The work is deciding which to sell, at what price, and when.

Track record

Eleven years building systems large operators run on.

Royal Caribbean

$3.5B

Cruise portfolio priced by a recommendation engine my data science team supported. 25,000+ price changes a week.

Universal Orlando

$400M+

Food, merchandise, and Express Pass revenue priced by a revenue management system I owned and rebuilt.

Revenue management

1–5%

Profit lift from the pricing programs above, varying by product, region, and season.

Workforce scheduling

10–15%

Lower labor cost from a scheduling optimizer I wrote for resort venues.

Results were measured by the employer at the time and describe roles I held. These companies are not clients of Larrea & Co.

Who it's for

Operators whose margins depend on these decisions, and the investors who own them.

Operators

Travel and leisure first: parks and attractions, resorts, ski areas, small-ship and river cruise, tour operators, marinas, and venues. Also airlines, logistics, retail, and healthcare or service businesses that schedule large workforces. Big enough that pricing and staffing decisions move real money; too small for a dedicated science team.

You sell perishable inventory or shift-based labor
A pricing, staffing, or allocation decision gets made every week
Your data exists, but nobody owns the optimization

Investors

Private equity firms and operating partners with leisure, hospitality, or capacity businesses in the portfolio, looking for value creation that repeats across assets.

Pricing discipline installed across sites, not one at a time
Labor productivity measured against a real baseline
Diligence on a target's pricing and capacity upside

What I don't do

Staff augmentation or filling a headcount gap
Platform migrations as the goal in themselves
Dashboards with no decision attached
Projects where no one owns the decision being improved

What I build

Four kinds of decisions, each built to fit the business.

Every engagement starts from a decision someone makes repeatedly. The method follows from the problem: an optimization model, a forecast, a simulation, an agent, or a redesigned process. Usually more than one.

Every system comes with the data pipelines, dashboards, and cloud architecture it needs to run in production.

Pricing & revenue management

Dynamic pricing grounded in measured elasticity, with an interface your analysts trust enough to use.

For example: a price recommendation engine with an override workflow, or an elasticity test before you commit to a new pricing strategy.

Capacity & allocation

Where to put ships, rooms, inventory, and demand, modeled as the constrained problem it is.

For example: a deployment model that weighs cannibalization, or a simulation of a new attraction's throughput before it opens.

Labor & scheduling

Schedules that cover forecast demand at the lowest cost while respecting the rules your people work under.

For example: a venue scheduling optimizer, or a peak-season staffing model that compares scenarios in minutes.

AI systems & process redesign

Agents and language models that work inside clear constraints, with people in control. The process gets redesigned first.

For example: an analyst copilot that explains price recommendations, or an agent that drafts group quotes within approved limits.

Also

Fractional data & AI leadership

For portfolio companies and mid-size operators without a head of data. A few days a month as your senior data and AI lead, on a monthly retainer.

Platform and architecture decisions, without a vendor's agenda
Oversight of data engineering and reporting work
An AI roadmap ranked by measurable value
Hiring and guiding your first analytics team

Why custom, why now

Custom used to be the expensive option. It isn't anymore.

A pricing engine or scheduling optimizer fitted to one company's constraints used to take a large team and a multi-year program. Mid-size operators bought generic software or went without.

AI-assisted development has changed that. A system built around your products, your rules, and your data is now a matter of weeks. It runs in your cloud and belongs to you.

What hasn't changed is the hard part: choosing which decision to model, measuring the lever honestly, and getting people to trust the result. That's what I've spent my career on.

How engagements will work

A defined decision, a measured result, and a system you own.

Frame the decision
Start from the choice you make every week and who owns the number, not from the data you happen to have.
Measure the lever
Elasticity, constraints, and a baseline come first, so the upside is a number before anything gets built.
Build the system
A working model and the interface your team will use, in your environment, on open-source foundations.
Hand it over
Source code, documentation, and training, so your team can run it without depending on me.

Fixed scope, fixed feePriced against a defined deliverable, not billed by the hour.

Diagnostic firstA short engagement sizes the opportunity and checks the data before any build.

Outcome-linked where it's measurableWhen lift can be measured cleanly, part of the fee can ride on it.

Selected work

Systems that changed a number someone was accountable for.

Work I led in corporate and consulting roles before Larrea & Co.

Royal Caribbean · Cruise

Pricing recommendation engine for a ~$3.5B portfolio

The problem. Analysts priced sailings largely by hand, with far more price decisions each week than they could review.

What I did. Led a team of three data scientists supporting the recommendation engine, and worked with IT on an analyst interface for inputs and overrides, so people stayed in control of the system.

2×

Sailings priced by the system doubled once analysts trusted the override path. Profit lift of 1–3% by region and season.

Universal Orlando · Resort

Revenue management across food, merchandise, and Express Pass

The problem. A proof-of-concept system took two to three hours per optimization run and limited how many price options could be considered.

What I did. Rebuilt it on Spark and SciPy, moved the Express Pass optimizer off SPSS, and prototyped knapsack-based pricing that handles non-linear and cross-product elasticity.

Hours → min

Optimization run time cut from hours to minutes. 2–5% yearly profit lift on $400M+ in revenue.

KPMG · Professional services

Engagement scheduling optimization for an audit practice

The problem. Peak-season staffing was assembled manually, concentrating workload into the worst months.

What I did. Built a scheduling optimization model and interface in Python and CVXPY on open-source solvers, so schedulers could compare scenarios instead of defending one plan.

16–23%

Lower peak-month utilization. Two follow-on engagements from the same client.

Writing

Notes on pricing, optimization, and applied AI.

The first pieces are in progress.

In progress · Pricing

The override button decides whether a pricing engine gets used

Why analyst trust, not model accuracy, set how much inventory our engine actually priced.

In progress · Revenue management

Measure elasticity before you build anything

What to test, and how, before committing to a pricing system.

In progress · Applied AI

Agents need constraints

Where language models fit in revenue management, and where a solver is still the right answer.

About

David Larrea

I'm an industrial engineer by training and have spent eleven years building analytics systems in production: pricing, labor scheduling, capacity planning, forecasting, and simulation.

At Virgin Voyages I was Director of Revenue Analytics & Data Science, leading the analytics workstream for a new cruise revenue management system. At Royal Caribbean I led a data science team supporting the pricing recommendation engine. At Universal Orlando Resort I owned the revenue management system for food, merchandise, and Express Pass, after starting there as an industrial engineer writing a labor scheduling optimizer and analyzing simulations for new parks and attractions. At KPMG I managed decision science engagements in optimization, simulation, and forecasting.

Today I'm a Solutions Architect at Databricks, helping enterprises design data and AI platforms. That keeps my view of what production AI really costs and requires grounded in practice. Larrea & Co. is independent of Databricks and recommends whatever platform fits the problem.

Currently: Solutions Architect, Databricks
Previously: Virgin Voyages · Royal Caribbean · Universal Orlando Resort · KPMG · Accenture
Education: M.S. Analytics, Georgia Tech
B.S. Industrial Engineering, University of Central Florida
Certifications: 10+ Google Cloud and Microsoft Azure certifications in data, ML, and architecture
Based in: Florida · remote or on site

Employers are named to describe experience only. They are not clients of Larrea & Co. and do not endorse it.

Stay in touch

Tell me about the decision you're trying to get right.

Whether you have a project in mind for later, want to compare notes, or would like new writing when it's published, send a note.

Email[email protected]