DIG-574 Product Definition: Mary’s questions through DV Digest
Product Definition
Template version: 1 | Created: 2026-10-27 UTC | The document Data Product Design (knowledge 36) produces, named the product definition so it is never confused with the design document (Ray, 2026-10-27)
| Product | Mary’s questions through DV Digest - effort DIG-574 - branch claude_ws/dig-574-arrival-and-identity |
| Governed by | Data Product Design (knowledge 36): the product stance (item 306), named releases (item 307), metric qualifier rows (item 308), the metric-era mart (item 309) |
| References | Questions predicted in the branch design (, Q-7 to Q-16); the connector, DV Digest (DIG-583); QA item ; the metric qualifier rows formerly in , retired into section 3 of this document; the metric-era work, DIG-587 |
| Author | Ray Serrano; edited by Claude - 2026-10-27 |
| Classification | Internal - Data Management |
| Official copy method | Instantiate this governed template through KNOWLEDGE_ITEM_FROM_TEMPLATE; retain the source template number and SHA-256 recorded by the procedure. |
How to complete this template
The product definition (in short, among the branch documents, the definition) decides what the product offers and how its offer reaches the consumer, before any table exists. It is the first of the branch documents: definition, profile, design, test plan, implementation guide, post-implementation guide. The others take their questions, metrics and releases from it (Ray, 2026-10-27). Work in order: the product and its consumer; the questions in the consumer’s words; the metric qualifier rows; the open points the product owners settle; the named releases; the grains and tables, one at a time. The template is elaborated through use: a section that does not fit a product says so with its reason, and the gap is fed back to knowledge 36.
1. The product
The consumer is Mary Mink, Director of Data Architecture and Governance at the distributor, asking from her own Claude through the DV Digest connector, which serves governed metrics only: a metric and its qualifiers. The decision the product serves: understanding what the distributor bought when it acquired Sunbelt Senior, how its business reaches the carriers, and how far the arriving data can be trusted.
Capabilities offered, by named release (section 5): the acquired book now; the history and origin of every record next; people across client numbers and agencies after that. Not offered yet: premium, which no source sends; and anything needing a second acquired agency, which the course’s source data does not yet include.
Name the consumer, the decision the product serves, and the capabilities the product owners judge to have business value. State what the product deliberately does not offer yet.
2. Consumer questions
| ID | Asked by | Asked at | Origin | Question in the consumer’s words |
|---|---|---|---|---|
| Q-7 | Mary Mink (predicted by Ray and Claude) | 2026-10-25 | Prediction, ratified as such (design ) | How many distinct people are our clients, across every agency we’ve acquired? |
| Q-8 | Mary Mink (predicted by Ray and Claude) | 2026-10-25 | Prediction, ratified as such (design ) | Which clients came to us through more than one agency? |
| Q-9 | Mary Mink (predicted by Ray and Claude) | 2026-10-25 | Prediction, ratified as such (design ) | Which policies does each client hold, with which carriers, in which roles (owner, insured, payor)? |
| Q-10 | Mary Mink (predicted by Ray and Claude) | 2026-10-25 | Prediction, ratified as such (design ) | How much business does each carrier write through us? |
| Q-11 | Mary Mink (predicted by Ray and Claude) | 2026-10-25 | Prediction, ratified as such (design ) | Which agent wrote each policy, and where do our sources disagree about it? |
| Q-12 | Mary Mink (predicted by Ray and Claude) | 2026-10-25 | Prediction, ratified as such (design ) | Which independent agents sell through us without an agency? |
| Q-13 | Mary Mink (predicted by Ray and Claude) | 2026-10-25 | Prediction, ratified as such (design ) | Where did this client’s record come from, and when did we first know it? |
| Q-14 | Mary Mink (predicted by Ray and Claude) | 2026-10-25 | Prediction, ratified as such (design ) | What did we know about this policy on a given date, and what changed since? |
| Q-15 | Mary Mink (predicted by Ray and Claude) | 2026-10-25 | Prediction, ratified as such (design ) | How many records arrive in a form we don’t accept, from which source? |
| Q-16 | Mary Mink (predicted by Ray and Claude) | 2026-10-25 | Prediction, ratified as such (design ) | When we acquire the next agency, how much of its book already overlaps ours? |
Use Q-n identifiers. A predicted question says so in its origin. Keep the consumer’s words exactly; parsing happens in section 3, never by rewording the question.
3. Metric qualifier rows
| Metric | Question | Qualifier |
|---|---|---|
| number of distinct people | Q-7 | person |
| number of distinct people | Q-7 | agency |
| number of agencies a client came through | Q-8 | client |
| number of agencies a client came through | Q-8 | agency |
| number of agencies a client came through | Q-8 | person |
| number of policies held | Q-9 | client |
| number of policies held | Q-9 | policy |
| number of policies held | Q-9 | carrier |
| number of policies held | Q-9 | role |
| total face amount (business written) | Q-10 | carrier |
| total face amount (business written) | Q-10 | distributor |
| total face amount (business written) | Q-10 | policy |
| number of policies written | Q-11 | agent |
| number of policies written | Q-11 | policy |
| number of policies written | Q-11 | source |
| number of agents named for a policy | Q-11 | policy |
| number of agents named for a policy | Q-11 | source |
| number of agents selling | Q-12 | agent |
| number of agents selling | Q-12 | agency affiliation |
| number of client records received | Q-13 | client |
| number of client records received | Q-13 | source |
| number of client records received | Q-13 | date first known |
| face amount | Q-14 | policy |
| face amount | Q-14 | status |
| face amount | Q-14 | date as of |
| number of changes | Q-14 | policy |
| number of changes | Q-14 | date of change |
| number of records rejected | Q-15 | source |
| number of records rejected | Q-15 | reason not accepted |
| overlap (aggregation open) | Q-16 | agency |
| overlap (aggregation open) | Q-16 | client |
| overlap (aggregation open) | Q-16 | policy |
One row per metric and qualifier it is asked by. A metric is one number the consumer asks for; a qualifier is a “by”, “for each” or “per” in the consumer’s words. These rows are the stored form: the matrix of metrics by qualifiers is a pivot generated from them for display, never stored or edited by hand.
4. Open points the product owners settle
| ID | Question | Open point | Decision | Reason | Decided by and when |
|---|---|---|---|---|---|
| OP-1 | Q-10 | How much business: its aggregation is not stated | Total face amount by carrier, the death benefit each carrier sends | Face amount is what the carriers send; premium is in no source, and adding it would be a later capability | Ray, 2026-10-27, answer 2y |
| OP-2 | Q-12 | Independent: no rule, and no source sends an agent’s affiliation | An agent named by a carrier’s policy data whom no acquired agency’s sales data names; in the course’s data, 17002001 | The rule reads affiliation from which sources name the agent, so no source has to send it | Ray, 2026-10-27, answer 3y |
| OP-3 | Q-7, Q-8, Q-16 | Person: no source sends one; client numbers are per agency | One later capability, people across agencies, release 3 | Needs the integration layer’s matching and a second agency in the source data | Ray, 2026-10-27, answer 4y |
| OP-4 | Q-16 | How much overlap: its aggregation is not stated | Open, settled when release 3 is designed | It depends on how people are matched | Ray, 2026-10-27, answer 4y |
An open point is anything the rows expose that a question leaves undecided: an unstated aggregation, a qualifier no source sends, a term without a rule. The product owners settle it and state why (item 306); it is not returned to the consumer as a blocker.
5. Named releases
| Release | Name: the idea that makes it cohesive | Questions | Why these together | Why not earlier or later | State |
|---|---|---|---|---|---|
| 1 | The Acquired Book: what was bought with Sunbelt Senior | Q-9, Q-10, Q-11, Q-15 | Clients, their policies, carriers and agents, and how clean the data arrived: one story about the acquisition | Its answers already exist and are checked; nothing in it waits on another capability | Expedited for time to market: merged into qa without the planned review (Ray, 2026-10-26); certified in QA (, gates 2 to 5); shipped to production on 2026-10-27 as v2026.10.1 (DIG-593; pull request #7 into main, squashed). Its tables, built before this method, are rebuilt under section 6 behind the same answers |
| 2 | What We Knew When: every record’s origin and history | Q-13, Q-14, Q-12 | Where each record came from, when it was first known, what it said on any date and what changed; independent agents are read from which sources report an agent | Q-12 placed here because building it into release 1 meant building it twice, once in the shape being retired | Next. Follows the full process release 1 skipped: its pull request is reviewed before it merges, and that review is part 1’s exam (Ray, 2026-10-27); it also remediates release 1’s design flaws |
| 3 | People, Not Numbers: the same person across client numbers and agencies | Q-7, Q-8, Q-16 | All three ask about people, not client numbers | Needs the integration layer and a second agency in the source data | Later |
Each release is coherent rather than comprehensive, weighed against time to market, and named for its idea (item 307). The calendar tag says when (knowledge 17, item 86); the name says what.
6. Grains and tables
| Grain: what one row represents | Metrics served | Qualifiers | Table | Diagram | Release | State |
|---|---|---|---|---|---|---|
| None designed yet | All | All | Tables are designed one at a time, each with its diagram (Ray, 2026-10-27, answer 5y); release 1 is served meanwhile by four views in MART built before this method (client_policy_fact, agent_policy_fact, policy_fact, finding_fact) | none | 1, 2, 3 | Not started |
Derived from section 3, never from the source’s or the vault’s structure: metrics needing the same grain share one table at that grain, named for it (item 309). Design one table at a time, each with its diagram; a row here is added when its table is designed, not in advance.
7. What the data must still provide
| Metric or qualifier | Questions | What is missing | Capability or change that provides it | Release |
|---|---|---|---|---|
| person | Q-7, Q-8, Q-16 | No source sends a person; client numbers are per agency | The integration layer’s matching of people, and a second agency in the course’s source data | 3 |
| agency affiliation | Q-12 | No source sends it | Derived by OP-2’s rule from which sources name each agent | 2 |
| overlap’s aggregation | Q-16 | Not decided (OP-4) | Decided when release 3 is designed | 3 |
A metric or qualifier with no source behind it becomes a later capability in a named release, not a silent gap.
8. Agreement checklist
| Requirement | Met, not met, or not applicable | Evidence or reason |
|---|---|---|
| Every question in the consumer’s exact words | Met | Section 2, copied from the design’s Q-7 to Q-16 |
| Every question has at least one metric qualifier row | Met | Section 3: 32 rows, 12 metrics, all ten questions |
| Every open point the rows expose is settled with its reason | Met, one deferred with its release | Section 4: OP-1 to OP-3 settled; OP-4 settled with release 3 |
| Every question in exactly one named release | Met | Section 5 |
| Every table traces to the rows and has its diagram | Not applicable yet | Section 6 fills one table at a time |
| Every gap names the release that closes it | Met | Section 7 |
Every question is in the consumer’s exact words; every question has at least one metric qualifier row; every open point the rows expose is settled with its reason; every question sits in exactly one named release or is stated as not offered; every table traces to the rows and has its diagram; every gap names the release that closes it.
9. Compliance: questions we propose
Added 2026-10-28 (Ray, answer 1y), from Claude’s research of 2026-10-27, at Ray’s direction: data architecture is scaffolding in the business’s eyes, while compliance is the heart of a director of data governance’s function, which exists to keep the company out of fines and lawsuits. These questions are ours, not Mary’s (numbered C-1 to C-8 since 2026-10-28, renamed from P-1 to P-8 so they never collide with a profile’s results, which the profiling standard numbers P-n): proposed to her, PROPOSED until she or Ray agrees each, and kept or dropped by what her use of them shows (the product stance, Pilot ). Every finding a detector writes cites the rule it breaks: the regulation, its section, the state’s version and its citation, kept as governed data with effective dates (Ray, 2026-10-27).
The pattern the research shows: the costliest failures are rarely dramatic fraud; they are ordinary data nobody connected: a consent no one could prove, a death the company knew of but did not act on, a replacement nobody counted.
| # | Proposed question | Why it matters (real cases) | What detects it | Data still needed | Release |
|---|---|---|---|---|---|
| C-1 | Which calls to prospects had no provable consent, or were made before it? | Telephone Consumer Protection Act suits are the industry’s busiest lawsuit engine: 507 class actions in the first quarter of 2025, up 112% on a year earlier, nearly 80% of TCPA suits filed as class actions. A court certified a class of up to $30 million against Liberty Mutual over leads bought from an aggregator; the FTC settled with Assurance IQ for $100 million and MediaAlpha for $45 million over deceptive lead generation | Each call joined to its consent record (lead seller, time, form); calls without one or before it | Lead, consent and call records | Release 2 (as of the call’s date) |
| C-2 | Which Medicare appointments lacked a documented scope of appointment, or came too soon after it? | CMS requires the beneficiary’s documented agreement to the scope of an appointment before it, and a 48-hour gap between the agreement and the meeting; the 2027 proposed rule would remove the 48 hours, so the rule in force on each date decides | Appointments against their scope agreements and the hours between, by the rule in force that day | Medicare appointment and scope-of-appointment records | Release 2 |
| C-3 | Which seniors’ policies were replaced soon after purchase, by which agents, with or without the replacement disclosure? | Rhode Island’s April 2026 consent agreement with Aflac covers replacements from 2022 to 2024 completed without the required internal replacement disclosure; California revoked an agent who repeatedly had seniors surrender annuities for new ones whose surrender fees ate their principal; Florida made twisting an annuity to a senior a third-degree felony | Replacements within a short window for the same client, by agent, client aged 65 and over, disclosure present or not | Policy history with replacements, client birth dates, disclosures | Release 2 (history), release 3 (the same client across agencies) |
| C-4 | Which premiums are paid from one account for unrelated clients, or by the agent, and which policies lapse right after the commission? | Two California agents submitted more than 600 fraudulent life applications, nearly $2 million in commissions, with premiums paid from the conspirators’ own accounts; a Texas agent was sentenced to 292 months on schemes paying 95 to 105 percent of first-year premium | Shared payment accounts, phones or addresses across unrelated clients; agent-paid premiums (rebating); early lapses after commission | Premium payment records, commission records, contact details | Release 3 (the same person, account or phone across records) |
| C-5 | Which clients appear on a deceased list while their policies stay in force with no claim? | Insurers used the Social Security Death Master File to stop annuity payments but not to find life policyholders who had died; MetLife paid nearly $500 million, and regulators reached multistate settlements with 27 life insurers | Clients matched to a deceased list against in-force policies without a claim; the same fact used for the company but not for the customer | A deceased list (invented, in the shape of the Death Master File) | Release 3 (matching people) |
| C-6 | Which sales were made by an agent not licensed, appointed or trained for that line, state and date? | NAIC Producer Licensing Model Act (a license required to sell, solicit or negotiate; appointment filed within 15 days, with states differing on sooner or later); NAIC Model #275 (a one-time four-hour annuity course, verified by the insurer before solicitation) | Each sale against the agent’s license, appointment and training as of the sale date | A licensing feed: licenses, lines, states, dates, appointments, training | Release 2 |
| C-7 | Which permanent life or annuity sales show money-laundering signs, and which agents sold them without anti-money-laundering training? | FinCEN’s rule for insurers (31 CFR Part 1025) covers permanent life and annuities, requires a written program integrating agents and brokers, their training, and suspicious activity reports | Third-party payers, early surrender of large single premiums, sales by untrained agents | Payment and surrender records, training records | Release 3 |
| C-8 | Which models and outside data sources touch a consumer decision, and have they been tested for unfair discrimination? | Colorado’s Regulation 10-1-1 (implementing SB21-169) requires life insurers using external consumer data to keep board oversight, an inventory, testing, monitoring and annual reporting, and does not carve out agents | An inventory of models and the outside data each uses: a data governance catalogue by another name | The model and data inventory | Outside the three releases; a governance catalogue |
What the releases turn out to be: release 2’s history (what was true on a date, by the rule in force on that date) serves C-1, C-2, C-3 and C-6; release 3’s people-matching serves C-3, C-4, C-5 and C-7. All three releases are compliance capabilities as much as analytics.
Sources (read 2026-10-27): TCPA filings, ActiveProspect (activeprospect.com/blog/tcpa-lawsuits-updates) and leadgen-economy.com; the Liberty Mutual certification, TCPAWorld, 2026-06-15; the FTC settlements, PossibleNOW (2025); CMS scope of appointment, the Oklahoma Insurance Department and CMS joint letter and Sanford Health Plan’s agent summary of the 2024 rule; the 2027 proposed rule, NABIP’s comment letter of 2026-01-26; Rhode Island DBR consent agreement INS 2025-IN-003 (2026-04-15); California Department of Insurance release 034-19; Florida SB 1372 summary (Colodny Fass); ICE and California DOI on the $1.9 million scheme; U.S. Attorney, Northern District of Texas, on the 292-month sentence; California DOI’s 2018 Death Master File report, Reuters on MetLife; NAIC Producer Licensing Model Act and its FAQ; NAIC Model #275 summaries (CFP Board, Senior Market Sales); FinCEN 31 CFR Part 1025 (govinfo.gov, FinScan); Colorado Division of Insurance on SB21-169 and Regulation 10-1-1.