Skip to content

FAQ

Short, accurate answers to the most common questions. Each links onward to the page with the full detail.

A full memo generation typically runs 15–25 minutes from the moment you trigger it. Document ingestion (parsing your PDFs, Excel model, and CoStar exports) is a separate, earlier step that usually completes in 2–5 minutes.

These are the headroom figures Memosa uses internally: a generation that is still running after 4 hours is treated as a dead worker and reclaimed, and document ingest stuck past 20 minutes is treated as failed. You will almost never see those limits — they exist to recover from crashes, not to cap normal runs.

See How It Works for the full pipeline.

Why is a number in the memo different from my model?

Section titled “Why is a number in the memo different from my model?”

Almost always because a higher-precedence source supplied the authoritative value. When two documents disagree, Memosa applies a fixed order — your direct input (100) beats the Excel model (90), which beats CoStar (70), which beats a PDF (50). So if your OM PDF says one cap rate and your Excel model says another, the Excel figure wins and the memo reflects it.

The memo footnotes the resolution, and the Canvas Provenance panel shows exactly which source won for each conflicted number. If the “winning” value still looks wrong, the issue is upstream — usually a parsing question. See Retrieval and Evidence and Diagnosing Problems.

Yes. Approval is not a one-way door for editing. A document goes DRAFT → EDITING → APPROVED, and the single backward step APPROVED → EDITING (reopen) is always available. Reopen, make your changes, and approve again — there is no limit on how many times you can do this.

What approval does is unlock exports (Investor Packet, Canvas PDF, OMCMS). See Document Lifecycle and The Approval Gate.

Do I have to approve the memo to export a PDF?

Section titled “Do I have to approve the memo to export a PDF?”

It depends on which export:

  • Canvas PDF of the document and read-only viewing are available without approval.
  • Investor Packet (DOCX/PDF) and OMCMS require the document to be in the APPROVED state. Attempting to generate an Investor Packet on an unapproved document returns a clear “document not approved” error rather than a partial export.

See The Approval Gate, Investor Packet, and Canvas PDF.

Access is governed by deal-level roles: owner > editor > commenter > viewer. You see and act on a deal according to your role, and cross-deal intelligence (comps, sponsor history, playbook) is scoped to your team — never shared into another team’s deals.

Collaboration is invite-based: you bring people onto a deal by inviting them at a specific role. See Roles and Permissions and Collaboration.

Upload the documents that carry the facts each section needs:

  • Excel model — pro-forma, cash flow, returns, debt schedule. This is the authoritative source for financial figures.
  • CoStar exports — market, demographics, rent and sales comparables.
  • PDF — the OM, sponsor materials, property description, legal/regulatory documents.

The web-chat upload accepts .pdf, .xlsx, .xls, and .csv , up to 50 MB per file and 20 files per upload .

A section can only be as good as its evidence — if you skip the CoStar export, the market section has nothing authoritative to cite. See Supported Documents and Retrieval and Evidence.

What do the [1] footnote markers in the memo mean?

Section titled “What do the [1] footnote markers in the memo mean?”

Every factual claim in the memo carries a citation that points to the exact source chunk it came from — a page of a PDF, a row of the Excel model, a CoStar table. The footnotes block at the end lists each source. This is how an analyst (or an IC) traces any statement back to a document. See Retrieval and Evidence.

A section says less than I expected — is something broken?

Section titled “A section says less than I expected — is something broken?”

Possibly. Thin output in a section almost always means thin evidence reached that section, and the usual cause is a missing or misclassified document. The most common case: CoStar reports that were uploaded but classified as generic PDFs, so the market section never saw them. Do not assume “the deal had no CoStar data” — verify your upload history first.

See Diagnosing Problems for the symptom-to-fix guide.

Yes. Memosa supports two intake paths: Slack (mention the bot in a deal thread and attach files) and web chat (a conversational intake inside the app). Both run the identical analysis pipeline. See Slack Intake and Web Chat Intake.

Where do I see the model’s underlying calculations?

Section titled “Where do I see the model’s underlying calculations?”

Memosa builds a dependency graph of your Excel model — when a driver cell changes, it can trace and re-flow every dependent number. In Canvas, the Metric Ripple feature lets you preview the downstream effect of changing an assumption before you commit it. See Metric Ripple.

Memosa degrades gracefully and surfaces actionable messages rather than stack traces. Start with Known Limits and Errors for what each behavior means and what to do, then Diagnosing Problems if you need to narrow down a quality issue.

  • src/canvas/services/web_intake_config.py — upload limits (max_file_size_bytes, allowed_extensions, max_files_per_upload), generation/processing stale thresholds.
  • src/canvas/models/section_lifecycle.pyDocumentState and DOCUMENT_TRANSITIONS (DRAFT → EDITING → APPROVED, and the reopen back-edge).
  • src/investor_packet/generation_validator.py_require_approved_state gating Investor Packet export on APPROVED.
  • src/intel/ontology/cre_ontology.yamlprecedence_weights (source-conflict resolution order).
  • src/canvas/auth/permissions.pyROLE_HIERARCHY (owner > editor > commenter > viewer).
  • Native memory: costar_classification_silent_misclassify.md (misclassified-CoStar failure mode), feedback_web_chat_has_excel.md.