# SR Automation — Meeting Brief

## What is this tool?

SR Automation is an Excel VBA macro running on a Windows VM. It is used by the Argentina team to process monthly subscriber reports from ~80 Polish cable/IPTV operators for Disney.

**How it works today:**
1. Polish operators send their monthly subscriber data via email (Excel/ODS/PDF attachment, subject: `RAPORTY`)
2. The macro connects to a generic Outlook mailbox, downloads attachments, and parses the data
3. Parsed data is compared against the Disney AIS historical data to detect discrepancies
4. Output files are generated per operator and saved locally
5. **Manual step:** results are uploaded to Disney's AIS system by a team member (Disney has not agreed to direct API write access — this step stays)

---

## Known Issues

### Tool / Technical Bugs

| # | Issue | Impact |
|---|---|---|
| 1 | Macro skips items during processing | Missing data in output, requires re-run |
| 2 | No crash recovery — re-running after a crash produces corrupted output | Team must know to close and reopen the file; if they re-run without doing so, garbage data is generated |
| 5 | Two clients sharing one email address — identification logic fails | Incorrect subscriber values assigned to wrong operator |

### Configuration & Process (Human Error)

| # | Issue | Impact |
|---|---|---|
| 3 | Mapping is maintained manually in an Excel sheet — additions, deletions, misconfigurations happen; some contracts have non-standard rules (e.g. client billed for only 80% of subscribers) | Wrong output values, requires manual correction |
| 6 | Team selects wrong reporting period | Entire run processes wrong month |

### Client Behavior (Outside Tool's Control)

| # | Issue | Impact |
|---|---|---|
| 7 | Some clients return the Excel template filled in as a scanned image or PDF instead of the original Excel file | Cannot be parsed automatically; requires manual handling |

### Already Resolved
- Rounding errors — fixed in current version

---

## VM Performance Problem

The macro runs on an Azure Virtual Desktop. Processing time for the **same dataset** varies significantly:
- Normal: ~10–15 minutes
- Bad days: 1–2 hours

The root cause is VM-level (Azure resource contention, CPU throttling) — not the macro logic itself. This has been getting worse over recent months and is the most visible pain point for the Argentina team.

**This cannot be reliably fixed within the current Excel/VM architecture.**

---

## Future Vision

Replace the Excel VBA macro + VM with a background service. The Argentina team would see ready data at the end of the month — no manual import, no waiting.

**Proposed stack:**
- **n8n** — workflow automation platform to orchestrate the end-to-end pipeline
- **Database** — store monthly extracts persistently; enables month-over-month history and audit trail
- **Microsoft Graph API** — connect to generic mailboxes via API, replacing the Outlook COM dependency

**What this solves:**
| Current problem | How new architecture addresses it |
|---|---|
| VM performance unpredictable | Service runs in cloud, no VM dependency |
| Re-run after crash corrupts data | Database-backed processing is idempotent — safe to re-run |
| Manual mapping in Excel prone to errors | Mapping stored in database, manageable via UI |
| Team selects wrong reporting period | Period auto-detected from email date |
| No history between months | Database retains all extracts by period |

**What stays the same:**
- Manual upload to Disney AIS (Disney requirement — no direct API access)
- Client-side issue with PDF/image submissions (process enforcement needed on operator side)

---

## Open Questions for This Meeting

1. **RPA vs n8n** — Argentina team was exploring RPA as an alternative. How does that align with the n8n direction? Is there a preference or constraint on tooling?
2. **Ownership model** — Jan has the domain knowledge; transformation team has the engineering capacity. What does collaboration look like?
3. **MVP scope** — what must work in v1 to replace the current tool for the Argentina team?
4. **Timeline** — when can the team realistically start? Are there dependencies (Azure AD app registration for Microsoft Graph, infrastructure provisioning)?
5. **Quick wins** — are there issues from the list above that can be patched in the current tool while the new service is being built?