Glossary — Tax Business KB

Quick lookups for the acronyms and terms used across this KB. Plain-English, one-liner definitions.

Business & metrics

• ARPU — Average Revenue Per User. Revenue ÷ users. The generic "how much is each customer worth per period" metric.
• ARPC — Average Revenue Per Customer. Same idea, per customer.
• ARPR — Average Revenue Per Return. Revenue ÷ tax returns filed — Intuit's tax-specific version (used in earnings, e.g., "ARPR +13% FY25"). In tax, ARPU ≈ ARPC ≈ ARPR since one person ≈ one return/year.
• Attach / attach rate. Extra products sold on top of the core purchase (audit defense, pay-with-refund); attach rate = % of customers who take them. Near-pure margin (file 05).
• CAC — Customer Acquisition Cost. Sales + marketing spend ÷ new customers won. Brutal in tax because all demand competes in the same ~10 weeks.
• LTV — Lifetime Value. Total revenue from a customer over the whole relationship (all years, all ecosystem products) — not one season.
• Mix shift. Changing the composition of customers (DIY → assisted) rather than the count — Intuit's core growth strategy.
• Free-to-paid conversion. % of free-tier users who end up paying (via complexity upgrades or add-ons). The economics of "free" depend on it.
• Churn / retention. Customers lost / kept year over year. Carryforward (below) is the big retention lever.
• Carryforward. Pre-filling this year's return from last year's data — a switching cost that keeps filers loyal.

Filing mechanics & regulation

• MeF — Modernized e-File. The IRS's XML-based electronic filing system; validates returns against schemas + business rules, returns Accepted/Rejected.
• ATS — Assurance Testing System. The IRS's annual acceptance testing that software must pass (per tax year) before it may e-file (Pub 1436).
• EFIN — Electronic Filing Identification Number. Issued to an Authorized IRS e-file Provider; required to e-file others' returns.
• ETIN — Electronic Transmitter Identification Number. Issued to software developers/transmitters for testing + transmitting.
• PTIN — Preparer Tax Identification Number. Required ID for anyone paid to prepare returns.
• EA — Enrolled Agent. Federally-licensed tax practitioner (IRS credential); with CPAs, the backbone of TurboTax Live's expert workforce.
• CPA — Certified Public Accountant. State-licensed accountant credential.
• WISP — Written Information Security Plan. The documented security program every tax preparer/filer must maintain (IRS Pub 4557 / FTC Safeguards Rule).
• §7216 (IRC Section 7216). The criminal statute barring tax preparers from using/disclosing tax return info beyond return prep without prior written consent (up to $1k fine / 1 yr per violation). The reason TurboTax shows separate "consent to use" vs. "consent to disclose" screens — and why TurboTax→Credit Karma sharing needs explicit consent (an affiliate is a separate entity under 7216).
• §6713. The civil twin of §7216 — $250 per unauthorized disclosure (cap $10k/yr), no intent required.
• Consent to use vs. consent to disclose. Two legally distinct §7216 consents: use = the preparer using tax data beyond prep (e.g., picking offers to show you); disclose = sending tax data to another entity (incl. Credit Karma).
• GLBA — Gramm-Leach-Bliley Act. The law making tax prep a "financial activity" subject to data-security rules; also governs credit data flowing into tax prep (the opposite direction from §7216).
• FCRA — Fair Credit Reporting Act. Governs credit-report data (Credit Karma's world).
• FTC Safeguards Rule. GLBA's implementing rule: MFA, encryption, pen-testing, incident-response, ~$50k/violation penalties.
• Free File / FFA — Free File Alliance. The 2002 IRS-industry deal: partners offer free filing; IRS (historically) stays out of the filing business.
• Direct File. The IRS's own free filing pilot (2024–25) — killed for FY2026.
• ALJ — Administrative Law Judge. An agency's in-house judge; the FTC's use of one against Intuit is what the 5th Circuit ruled unconstitutional (file 10).
• IP PIN — Identity Protection PIN. IRS-issued PIN preventing identity-theft refund fraud.

AI / technology

• NLP — Natural Language Processing. The AI discipline of understanding/generating human language (intent detection, document reading, summarization). The older umbrella term; LLMs are the current dominant technology for it — "NLP models" often signals older task-specific systems vs. the GenOS/LLM stack.
• LLM — Large Language Model. One big general model (Claude, GPT) that handles most language tasks without task-specific engineering; what collapsed the tax-knowledge moat (file 11).
• ML — Machine Learning. The broader umbrella: models that learn patterns from data (includes NLP, matching/routing models, fraud detection).
• Knowledge engineering. Encoding domain rules (the tax code) into machine-checkable form — the pre-LLM heart of TurboTax's engine; still the accuracy backstop.

Data & platforms

• Data lake. One central repository holding raw data of every shape, cheaply (files in object storage). Great for scale; without curation it degrades into a "data swamp" — hard to trust or query.
• Data warehouse. The opposite trade: structured, curated, fast-SQL analytics (e.g., Redshift) — reliable but rigid, and poor with raw/unstructured data.
• Lakehouse. The merger: lake-cheap storage underneath, warehouse-grade features on top (ACID transactions, schemas, fast SQL) — so ONE copy of the data serves dashboards, data science, ML, and streaming. Intuit runs on one (Databricks + Redshift on the same data).
• ACID transactions. Database-grade guarantees (atomic, consistent, isolated, durable) so concurrent reads/writes don't corrupt or half-apply — what lakes lacked and lakehouses added.
• Data mesh. An organizational architecture, not a technology: instead of one central data team owning everything (a bottleneck that doesn't understand the domain data), each domain team owns its data as a product — with quality SLAs, documentation, discoverability — on shared self-serve infrastructure with federated governance. Intuit moved lake → mesh. (ECC's Tier 0/1/2 SLAs are data-as-a-product thinking in action.)
• Spark / Spark Streaming. The dominant open-source engine for large-scale data processing; the Streaming variant processes events continuously as they arrive rather than in nightly batches.
• Flink. Another streaming engine, built real-time-first; used where low-latency event processing matters.
• Batch vs. streaming. Batch = process accumulated data on a schedule (nightly job); streaming = process each event as it happens. Season-critical metrics want streaming; deep history wants batch.
• Databricks. The commercial platform built around Spark (notebooks, SQL, ML tooling) — one of the two query front doors to Intuit's lakehouse.
• Redshift. AWS's SQL data warehouse — the other front door; where the ECC re-platform landed.
• Feature store. A serving layer that keeps ML-model inputs ("features," e.g., an expert's current capacity) fresh and consistent for real-time decisions like VEP matching.
• Event stream / journey timeline. An append-only log of things that happened (page views, filings, calls) per customer — what powers "summarize this member's journey."

Intuit-specific

• VEP — Virtual Expert Platform. Intuit's platform connecting customers to AI + human experts; powers TurboTax Live / QuickBooks Live.
• GenOS. Intuit's internal Generative AI Operating System (CTO Balazs architected; Chief AI Officer Srivastava scaled).
• Intuit Assist. The GenAI financial assistant embedded across Intuit products.
• Live / Live Assisted / Live Full Service. TurboTax's expert tiers: on-demand expert help vs. an expert doing the whole return.
• CG — Consumer Group. The Intuit org spanning TurboTax + Credit Karma (Mark Notarainni, EVP/GM). Note: Intuit's reported "Consumer" segment (~$4.9B FY25) is TurboTax only; Credit Karma reports separately.
• CK Money — Credit Karma Money. CK's banking product; where Refund Advances land (the TurboTax→CK bridge).
• Prosperity ecosystem. Intuit's framing for the connected TurboTax↔Credit Karma↔QuickBooks↔Mailchimp flywheel.

Seasonal / operational

• The season. Late Jan (IRS opens) → Apr 15 (deadline), plus the Oct 15 extension bump — where most revenue happens.
• Change freeze / freeze window. A dated period when no changes are pushed to production (peak season protection).
• Peak days. The highest-load filing days (just before deadlines).