Solution Knowledge Transfer Plan

Status: In Progress Last Updated: 2026-01-24 Source Documents:

• new_employee_back-office_orientation.pdf
• Data Science Confluence exports (6 PDFs)

Completed Work

Documents Created

• docs/solution-overview.md - Company identity, core concepts, verticals
• docs/path2acquisition-flow.md - 11-stage data flow breakdown
• docs/glossary.md - 100+ term definitions (expanded with DS terms)
• docs/tools-and-systems.md - Tools, security, Agile process
• docs/bert-overview.md - BERT unified backend platform
• docs/data-science-overview.md - XGBoost modeling, P2A3XGB pipeline, DS team operations

Integration

• Updated CLAUDE.md with solution overview section
• Updated README.md with solution knowledge section
• Cross-referenced existing company docs
• Added Data Science terms to glossary (ARA, AUC, AUCPR, DRAX, Encoded Windowed Variables, ERR, FTB, Future Projections, Hotline Scoring, MGR, MPT, P2A3XGB, PBT, Responder Interval, Windowed Variables, windowedEndDate, XGBoost)

Next Session: Deep Dive Topics

The following areas from the Page 2 diagram require deeper exploration and documentation:

1. BERT System

Priority: High Status: Partially documented (see docs/bert-overview.md) Known: BERT = Base Environment for Re-tooled Technology - unified backend application platform consolidating Order App, Data Tools, Dashboards, Response Analysis, and other tools. Named August 2023 (submitted by Wells). Questions to answer:

• Current state of BERT development and module integration
• What are the inputs and outputs for each module?
• How does it generate “Variables” for modeling?
• Architecture details (serverless migration status)
• Is there existing technical documentation?

2. Modeling Process

Priority: High Status: ✅ Documented (see docs/data-science-overview.md) Documented:

• XGBoost classification model (P2A3XGB) for Path2Acquisition
• 6-step pipeline: Train Select → Train → Score Select → Score → Reports Select → Reports
• Variable types: Profile, Windowed, Encoded Windowed
• Model quality assessment: PBT graphs, Model Grade Report (MGR)
• Future Projections, Hotline Scoring, DRAX reports
• Team: Igor Oliynyk (Lead), Morgan Ford, Erica Yang Remaining questions:
• How are Affinity models different from Propensity models?
• Score File structure details

3. Synergistic Titles

Priority: High Questions to answer:

• How is “synergy” between titles determined?
• Is this algorithmic or manual selection?
• What data drives synergistic title identification?
• How does this feed into Preselect?
• Example: McGuckin Hardware → Home Depot, Lowes (why these?)

4. Match Back Process

Priority: Medium Questions to answer:

• How are responses linked back to original mailings?
• What is the Keycode structure?
• How long is the match-back window?
• How are “Lists of Lists” and “Multis” handled in attribution?
• What systems perform match back?

5. Digital Solutions Branch

Priority: High (strategic importance) Questions to answer:

• How does the Householding → Digital Solutions path work?
• How does this connect to LiveRamp integration?
• What are “License Files”?
• How do Hashed Emails (HEM) flow through to digital activation?
• What is the relationship between Site Visitors and digital audience products?

6. Data Flow Details

Priority: Medium Questions to answer:

• Complete vs Incremental transactions - when is each used?
• CASS/DPV/DPBC/NCOA - exact sequence and purpose
• How does JSON transaction storage work?
• What is the Title Key and how is it used?
• BIGDBM integration - what data comes from them?

7. Preselect & Universe

Priority: Medium Questions to answer:

• How is the Universe defined for a campaign?
• Responder vs Prospect - how are these distinguished?
• DNM vs DNR - business rules
• Omits & Suppressions - categories and sources
• What determines Merge Cutoff timing?

8. Service Bureau & Fulfillment

Priority: Low Questions to answer:

• Which service bureaus does P2R work with?
• Merge/Purge process details
• How are Hits and Nets reported?
• Computer Validation specifics
• Mail File format requirements

Approach for Next Session

1. BERT - ✅ Partially documented (see docs/bert-overview.md). Continue documenting module details.
2. Modeling - ✅ Documented (see docs/data-science-overview.md). Minor questions remain.
3. Then Synergistic Titles - Key to understanding audience selection
4. Then Digital Solutions - Strategic priority for 2025/2026
5. Remaining topics as time permits

Resources Needed

• Access to technical documentation (Confluence DEV space)
• Possibly conversations with:
  • Engineering (John Malone) - BERT, Data Tools
  • Data Science (Igor Oliynyk) - Modeling
  • Operations - Match Back, Service Bureau

Session Prompt

Use this prompt to resume tomorrow:

Continue the Path2Acquisition solution knowledge transfer. We've created initial documentation from the new_employee_back-office_orientation.pdf. Now we need to deep dive into:

1. BERT system - what it does, inputs/outputs
2. Modeling process - Affinity vs Propensity, how models are built
3. Synergistic Titles - how synergy is determined
4. Digital Solutions branch - connection to LiveRamp
5. Match Back process - linking responses to mailings

Start with BERT. Check docs/solution-knowledge-transfer-plan.md for the full plan and questions to answer. Use Confluence/Jira to find technical documentation where available.