George Nakhleh · Analytics as a Service, as Products, as a Platform

Range

Three ways I deliver analytics

Most analysts live in one of these. Service is the judgment I bring to a single decision. A product turns that judgment into something durable the org keeps using. A platform lets the whole org build and trust its own analytics on top. Each tier lifts the next: service shows what's worth building, products show what the platform has to guarantee, and the platform multiplies what everyone above it can do.

01 · Service

In the room

Judgment delivered straight to a decision. I build the forecast, the model, the evaluation, and I'm in the meeting where it turns into committed budget and headcount. It's also where I see what's worth making permanent.

The capacity model that sets how a fast-growing flexible workforce gets staffed each year, the supply-chain value analysis that turned benchmarking into validated contract savings, the impact reviews that tell leadership which programs to keep funding.

02 · Products

Outlasts the ask

When a decision keeps coming back, the answer should be a durable, owned product that keeps delivering: dashboards, data marts, chatbots, a reusable AI toolkit, each with an owner, real users, and a roadmap that keeps it current as the business moves.

A tiered care-team performance product that carries one case-mix-adjusted methodology from the leadership view down to the individual, plus the support reporting and chatbot the operations org reaches for every day.

03 · Platform

Self-serve, still canonical

The layer that lets the whole org build and trust its own analytics: governed definitions, lineage, promotion paths, evals. Most teams stop at products, and a stack of products is not a platform.

The company's first governed semantic layer, so non-technical leaders query in plain English against one set of definitions, with promotion paths that graduate a strong local metric into the org standard and telemetry that retires the dashboards no one opens.

Point of view

How I think about this work

Bedrock and voice

AI has collapsed the cost of producing a chart to nearly zero, which is great for speed and terrible for trust. The answer is governed bedrock underneath: deterministic, version-controlled metric definitions every tool and every AI reads from, and a consumption surface that earns attention through rigor and disclosure.

Democratization isn't automation

The lazy AI playbook automates the analyst and hands the domain expert a chatbot. The real win is agency: the people closest to the work can inspect, question, and create with the information that describes it. Engineers maintain the scaffolding; domain experts drive the content.

Standards as guardrails, not walls

Centralize everything and people build shadow metrics; democratize everything and coherence dies. The third option: endorsed standards that are visible and voluntary. You can depart from them; the system just knows, shows it, and surfaces the patterns so good variants graduate into the standard.

The reading list behind it Click to openClick to close

A longer-running reading project in cybernetics, measurement theory, and organizational learning. Lead with Moneyball; defend with the canon. Two working questions per book, not summaries. Happy to talk about any of it.

Information & the economics of knowledge

Shannon, A Mathematical Theory of Communication (1948)
Wiener, The Human Use of Human Beings (1950)
Anderson, The Long Tail (2006)
Handy, We the Purple People (2021)

The measurement tradition

Taylor, Principles of Scientific Management (1911)
Deming, The New Economics (1993)
Ohno, Toyota Production System (1978)
Grove, High Output Management (1983)
Kaplan & Norton, The Balanced Scorecard (1992)
Austin, Measuring and Managing Performance in Organizations (1996)
Muller, The Tyranny of Metrics (2018)
Porter, Trust in Numbers (1995)
Ackoff, Redesigning the Future (1974)

Cybernetics & systems

Wiener, Cybernetics (1948)
Ashby, An Introduction to Cybernetics (1956)
Bateson, Steps to an Ecology of Mind (1972)
Beer, Brain of the Firm (1972) · Designing Freedom (1974)
Meadows, Thinking in Systems (2008)
Sterman, Business Dynamics (2000)

Optimization & decision

Bertsimas & Tsitsiklis, Introduction to Linear Optimization (1997)
Boyd & Vandenberghe, Convex Optimization (2004)
Ben-Tal, El Ghaoui & Nemirovski, Robust Optimization (2009)
Goldratt, The Goal (1984)
Hubbard, How to Measure Anything (2007)
Pearl, The Book of Why (2018)

Institutions, sensemaking, failure

Argyris & Schön, Organizational Learning (1978)
Douglas, How Institutions Think (1986)
Perrow, Normal Accidents (1984)
Weick, Sensemaking in Organizations (1995)
Davies, The Unaccountability Machine (2024)

Health systems & care

IOM, Best Care at Lower Cost (2013)
Wachter, The Digital Doctor (2015)
Berwick, Escape Fire (2002)

Technology, scale, autonomy

Illich, Tools for Conviviality (1973)
Alexander, A Pattern Language (1977)
Sutton, The Bitter Lesson (2019)
Carse, Finite and Infinite Games (1986)
Weil, The Need for Roots (1943)

Cultural shorthand

Lewis, Moneyball (2003)

Experience

Where the work happened

Hinge Health2022 – PRESENT

Senior Data Analyst, Operations Analytics (Analytics Engineering & AI Platform)

REMOTE

Platform building, decision support, and AI enablement across support, care, and clinical operations, working with operations leadership, finance, data engineering, quality, and clinical teams.

Built my analytics team's AI harness: a shared repo of reusable skills, context, and patterns (primary author, 57% of commits), with data guardrails and a cost-accountability framework. Paired it with the company's first semantic metric layer, so non-technical leaders query governed data in plain English with no ticket, at 16–22× lower cost per query. Wrote the onboarding guide now cited alongside IT and Security docs, and coached no-Git staff to self-serve safely.
Modernized support analytics end to end: consolidated disparate reporting into the company's first unified, production-grade platform in dbt and Databricks (email, chat, and phone for a 124-agent org), with the data engineering to make core operations modelable, unified measurement standards, the org's first individual performance reporting, and a support chatbot. Usage telemetry then consolidated or retired over a dozen legacy reports.
Owned demand forecasting and capacity planning for 1+ years: the model that sets annual staffing and ongoing schedules for a flexible workforce at a company growing over 10% a year. Consolidated fragmented forecasts, encoded business heuristics where no history existed, and built the reporting front end. A capacity-stress metric made a workload surge visible (117 to 290 units per worker per week at flat headcount), and I reconciled a 34-point utilization gap between two teams measuring from different systems.
Designed and owned a tiered care-team performance product for 1+ years: leadership, manager, and individual views, coherent through dbt, replacing fragmented team-by-team reporting with one holistic methodology the people being measured can interrogate. Made the call on case-mix adjustment so workload difficulty doesn't distort evaluations, and landed it across heavily invested stakeholders through iterative change management, not by fiat.
Run the team's innovation-lab work: mapped 20 compliance-sensitive metrics across five regulated domains and proposed a composite "Regulated Risk Signal" executive metric; ran a self-initiated, pre-registered causal study (n=12,197) connecting care-team quality and efficiency to member engagement and outcomes, showing a +13.4pp lift; and built novel information experiences, from a "Wordle for data" quarterly KPI celebration to animated operational flow visualizations.
Built a full experimentation stack for insurance verification: two sequential A/B tests in Statsig, attribution, an SMS launch, and a conversion-likelihood scoring model so outreach works from a prioritized list. Represented analytics on the Tableau to Mode BI migration.
Introduced Jira to the team and later led the migration to Linear, architecting the program structure (98 issues across 7 cycles) and deploying multi-agent AI infrastructure where agents create, label, and route follow-on work under a self-governance framework.

MD Anderson Cancer Center2021 – 2022

Supply Chain Decision Support Specialist (promoted)

HOUSTON, TX

Value analysis and contract decision support at a top US cancer center, working with clinical, supply chain, and finance leaders.

Lead analyst on the Cost of Care initiative: found high-impact savings in medical-supply and drug utilization with Vizient benchmarking, led contract savings validation that secured $1M+ in documented improvements, and set the department's supply-chain analytics strategy roadmap, presenting findings to supply-chain and finance leadership.

MD Anderson Cancer Center2020 – 2021

Supply Chain Decision Support Analyst