The Alternative-Investment-Technology Glossary
The vocabulary at the intersection of finance and engineering — defined plainly, for the investors and operators who have to make decisions with it.
Technology Due Diligence
Technology due diligence
Technology due diligence is the independent assessment of a target company's technology before an investment or acquisition. It covers codebase quality, infrastructure and scalability, security, technical debt, AI readiness, and the engineering team, then translates each into investment-grade findings: the risk, how much it matters, and what it will cost to fix.
In alternative investing: For PE and VC investors, it answers whether the technology will scale, what remediation will cost, and whether the team can deliver — before capital is committed.
Technology Due Diligence →Investment-grade technical findings
Investment-grade technical findings are diligence conclusions written for an investment committee rather than for engineers. Each finding states the risk, its materiality to the thesis, and a quantified remediation cost — so a partner who doesn't read code can act on it.
In alternative investing: It's the difference between a report that lists technical observations and one that tells you how a risk affects the deal and the model.
Technology Due Diligence →AI readiness (real AI vs. an AI slide)
AI readiness assessment determines whether a company's AI is a genuine, defensible differentiator or a thin wrapper on a third-party model. It examines proprietary data, model differentiation, evaluation discipline, deployment maturity, and key-person risk — the signals a pitch deck doesn't show.
In alternative investing: In diligence it answers the now-standard question: is this an AI company, or a company with an AI slide?
Technology Due Diligence →Key-person risk
Key-person risk is the exposure created when critical knowledge — how a system works, why it was built a certain way, where the data quirks are — lives in one or two people's heads rather than in documentation or code. If they leave, the capability leaves with them.
In alternative investing: In a target or a roll-up integration, undocumented institutional knowledge is a quantifiable risk to value, not just an HR concern.
Technology Due Diligence →Fractional CTO
A fractional CTO is an experienced technology executive who works with a company part-time — typically a few days a month — instead of as a full-time hire. They provide senior judgment on strategy, architecture, vendor selection, and where AI fits, without the salary, equity, and recruiting cost of a full-time executive.
In alternative investing: It fits emerging managers and portfolio companies that need executive-level technical judgment for specific decisions, not a forty-hour week.
Technology Leadership →Data & Infrastructure
Data warehouse
A data warehouse is a central system that ingests data from a firm's many source systems, conforms it to one shared schema, and exposes it in a clean, query-ready form. It becomes the single source of truth every downstream report and dashboard reads from.
In alternative investing: For a firm or platform whose numbers never agree, the warehouse is the layer that makes reports reconcile because the data underneath finally does.
Data Warehouse Design & Implementation →Medallion architecture (bronze / silver / gold)
Medallion architecture organizes a data warehouse into three progressive layers. Bronze is the raw, unmodified copy of source data; silver is validated, deduplicated, and conformed to a shared schema; gold is the business-ready layer of pre-calculated metrics that analysts and dashboards actually use.
In alternative investing: In a PE roll-up it creates a repeatable path for every acquisition's data: land in bronze, conform in silver, surface as platform metrics in gold.
Data Warehouse Design & Implementation →The N-systems problem
The N-systems problem is what happens when a platform is assembled from many companies that each arrive with their own ERP, CRM, and definitions of basic concepts like customer, job, and revenue. Complexity grows multiplicatively, not additively, so the combined platform can't answer basic questions without manual spreadsheet stitching.
In alternative investing: It's the defining data challenge of PE roll-ups, and the reason board questions stay unanswerable for the first 12–18 months unless a consolidation layer is built early.
Data Warehouse Design & Implementation →Entity resolution
Entity resolution is the process of recognizing that records in different systems refer to the same real-world entity — that "Johnson & Sons HVAC" and "Johnson and Sons Heating & Cooling LLC" are the same company — and merging them into one golden record. Modern approaches combine deterministic matching (exact keys) with probabilistic and AI matching for the fuzzy cases.
In alternative investing: In a roll-up it's how you answer strategic questions like total wallet share with a customer that several acquired companies all serve; AI handles the bulk, with a confidence threshold routing ambiguous matches to human review.
Data Warehouse Design & Implementation →Master data management (MDM)
Master data management is the discipline of maintaining one authoritative, deduplicated record for each core entity — customer, vendor, employee, product — across all of a firm's systems. It resolves overlaps and conflicting definitions so analysis isn't double-counting or comparing apples to oranges.
In alternative investing: It's where roll-ups should start (usually with customers), because that's where the strategic value of consolidation is highest.
Data Warehouse Design & Implementation →ETL / ELT
ETL (extract, transform, load) and ELT (extract, load, transform) are the two patterns for moving data from source systems into a warehouse. ETL transforms data before loading it; ELT loads raw data first and transforms it inside the warehouse, which is the more common modern approach.
In alternative investing: These pipelines are the plumbing that keeps a consolidated reporting layer fed and current as new portfolio companies and data sources come online.
Data Warehouse Design & Implementation →Data quality & governance
Data quality and governance is the set of validation, reconciliation, lineage, and ownership practices that make data trustworthy. Quality checks at the point of ingestion flag anomalies before they pollute downstream reports; lineage proves where every number came from.
In alternative investing: A single bad board report from one add-on's dirty data can make a CFO abandon dashboards for spreadsheets — governance is what prevents that loss of trust.
Data Warehouse Design & Implementation →Self-serve analytics
Self-serve analytics lets non-technical users explore data and answer their own questions through dashboards and analytics portals, without filing a request to an analyst. The reporting layer is designed so tomorrow's question doesn't require a new build.
In alternative investing: It frees the investment team and portfolio-company managers from queueing behind a single analyst for every new cut of the numbers.
Business Intelligence & Reporting →AI
Retrieval-augmented generation (RAG)
Retrieval-augmented generation is an architecture where a large language model answers from a specific body of documents by first retrieving the relevant passages and then generating an answer grounded in them. It produces cited, source-grounded answers instead of relying solely on the model's training.
In alternative investing: It's the foundation for asking plain-English questions over a firm's private knowledge and getting answers that point back to the source documents.
AI Integration & Custom Software →Data-room RAG
Data-room RAG applies retrieval-augmented generation to a deal or diligence data room, letting a team query thousands of documents in plain English and get grounded, cited answers. It accelerates document review without the model inventing facts that aren't in the room.
In alternative investing: For diligence and portfolio monitoring, it turns a data room from a manual reading exercise into a queryable knowledge base — with the citations an IC needs to trust the answer.
AI Integration & Custom Software →AI agent / agent pipeline
An AI agent is a system that uses a language model to carry out multi-step tasks — calling tools, retrieving data, and chaining actions toward a goal — rather than just answering a single prompt. An agent pipeline composes several of these steps into a repeatable workflow.
In alternative investing: In practice this means automating multi-step research and operations work, like gathering and synthesizing inputs for a diligence memo, with the work integrated into existing systems.
AI Integration & Custom Software →Build vs. buy (AI)
Build vs. buy is the decision of whether to build a capability in-house or adopt an off-the-shelf product. For AI, off-the-shelf tools fit the average customer; building or integrating fits when the workflow, data, or security requirements are specific enough that a generic tool can't express them.
In alternative investing: For investment firms the calculus turns on data privacy, fit to a specialized workflow, and whether you want to own the capability or rent it — a senior, independent judgment call.
Technology Leadership →Research Methods
Alternative data
Alternative data is non-traditional data used to inform an investment thesis or monitor a company — card and transaction panels, web-scraped pricing and reviews, geospatial and foot-traffic data, satellite imagery, and public government datasets, among others. It spans dozens of categories at prices from free to over a million dollars a year.
In alternative investing: The value is in sourcing the signal that answers your specific question, cleaning it, and turning it into something defensible — not in buying the most expensive feed.
Bespoke Data Analysis →Control group
A control group is a comparison case that was not exposed to the factor you're studying but was subject to the same broader conditions. If the outcome changes in the exposed group but not the control, the factor — rather than the shared conditions — is the likely cause.
In alternative investing: For an alt-data backtest, a control group is how you show a signal reflects the effect you claim and not a market-wide move that would have happened anyway.
Bespoke Data Analysis →Natural experiment
A natural experiment uses an external, unplanned change — a rule change, a policy shift, an exogenous shock — as if it were a controlled experiment, because the change wasn't chosen by the subjects being studied. It lets you test causation when a true randomized experiment isn't possible.
In alternative investing: In research it's how you move a claim from "these two things correlate" to "this change caused that outcome" — the standard a thesis should clear before capital follows it.
Bespoke Data Analysis →Confounder
A confounder is a third variable that influences both the cause and the effect you're studying, creating a correlation that isn't causal. Sound analysis identifies and rules out plausible confounders before concluding that one thing drives another.
In alternative investing: Failing to rule out confounders is how a backtest finds a signal that evaporates in production; eliminating them is what makes a finding trustworthy.
Bespoke Data Analysis →Ecological fallacy
The ecological fallacy is the error of assuming that a relationship seen at the group or aggregate level also holds for individuals within it. A strong correlation across eras or markets can sit alongside a near-zero correlation within any single one.
In alternative investing: It's a common trap in quant research: a cross-sectional pattern and a time-series pattern can point in opposite directions, and conflating them produces a signal that doesn't trade.
Bespoke Data Analysis →