Deal Pipeline

The memory of the search. Every conversation, every stage change, every next action, in one place you can actually read.

Proprietary deals are lost to forgotten follow-ups far more often than to losing bidders. The owner who said "call me in the spring" and never got the call sells to whoever did.

Step 0 — Load

1. Read thesis-profile.md for buyer identity and size box. If it does not exist, stop and walk the user through creating it from templates/thesis-profile.md.
2. Read pipeline.csv in the workspace. Create from templates/pipeline.csv if absent.

date_added,company,contact_name,title,channel,sender,stage,
first_touch_date,last_touch_date,touch_count,reply_class,
next_action,next_action_date,valuation_discussed,probability,
warm_path_via,nda_status,notes,updated_date

One row per conversation, not per company. A broker taking over an owner thread is a new row, because it is a new conversation with its own state.

Stages

A conversation moves forward or it closes. It does not sit in limbo without a next action.

Rules

1. Every row in an active stage needs a next_action and a next_action_date. No exceptions. A conversation without a next action is a conversation you have already lost, you just do not know it yet.
2. Do-not-contact is permanent. A not-selling row flagged opt-out is never revisited, on any channel, ever.
3. dormant needs a real revisit date, taken from what the owner actually said. "After tax season" becomes a date.
4. Never delete a row. Close it with a reason. In a multi-year search, the history is the value.
5. Log the touch when it happens, not at the end of the week.

Logging a conversation

When the user describes a call or reply, capture:

• What was actually said, in the owner's words where it matters
• Any number discussed, and who said it. An owner's "I'd want at least 8x" is their statement, not a valuation
• What they asked for, and what you promised, by when
• NDA status and anything confidentiality touches
• The next action and its date

Note discipline

Notes get read back later, sometimes by partners, lenders, or counsel, and occasionally by the owner if a thread gets forwarded.

• Write facts, not interpretation. "Said revenue is around $9M" not "doing well"
• valuation_discussed records only numbers actually spoken, with who said them. Never your estimate — estimates live in targets.csv, labeled with their basis
• Never speculate about an owner's health, family, finances, or motives
• Quote commitments exactly. "I'll think about it" is not "interested in our LOI"

See references/compliance-and-ethics.md on personal data and confidentiality hygiene.

Probability

Keep it honest and coarse. Precision here is false precision.

A signed LOI is not a deal. Diligence kills more deals than negotiation does. Do not let one live LOI talk you into stopping top-of-funnel work.

Daily view

On request, or when the user asks what needs doing:

TODAY — 4 August

Overdue (3):
  [Company A]  — promised the buyer one-pager 4 days ago. Still not sent.
  [Company B]  — follow-up was due 1 August
  [Company C]  — call was 6 days ago, no recap sent

Due today (2):
  [Company D]  — send NDA draft they asked for
  [Company E]  — owner said "call after the trade show" (ended Friday)

Awaiting their reply (5):
  [Company F]  — 8 days since one-pager sent
  ...

This month's dormant revisits (2):
  [Company G]  — owner said "after my season ends in September"
  ...

⚠ 2 rows in an active stage have no next action:
  [Company H], [Company I]  — set one or close them

Lead with what is overdue. That is the whole point of the view.

Writing changes

🚧 HARD GATE — before writing

Show every proposed change to pipeline.csv before writing it: new rows, stage transitions, and any row being closed, with the reason on each. Flag any stage jump that skips nda-signed before an IOI, and any closed row — confirm signing and funding actually happened before it is written. Get approval. Then write, and update updated_date on every touched row.

Weekly summary

PIPELINE SUMMARY — week ending 1 August

Thesis:      HVAC services, $1M to $3M EBITDA, Southeast US

Funnel:
  identified       184
  qualified        118
  contacted         71
  in-dialogue        9
  nda-signed         3
  ioi-sent           1
  loi-sent           0
  diligence          0
  closed             0
  dormant           14   ← nearest revisit: September (2 rows)
  not-selling       12   (1 opt-out, permanent)
  passed             5

Movement this week:
  +2 in-dialogue      [Company A], [Company B]
  +1 nda-signed       [Company C]
  +1 ioi-sent         [Company D]  (range shared, awaiting reaction)
  +3 dormant          revisit dates set

Conversion:
  contacted → in-dialogue   13%
  in-dialogue → nda-signed  33%
  nda-signed → ioi-sent     33%

Top of funnel is the constraint. One IOI out is not a pipeline. At current
conversion, one closed deal needs roughly 500 contacted; 71 are contacted
and discovery has stalled for two weeks.

Recurring pattern:
  "Talk to my broker"   3   ← brokered processes, decide policy once

Overdue actions: 3
Needs your attention: [specific list]

Name the constraint. A summary that does not say what is actually blocking the search is a status report, not a useful one.

Notes

• Update it the same day. A pipeline reconstructed from memory on Friday is fiction.
• The overdue list is the most valuable output. Show it first, every time.
• Do not inflate probability. A weighted view that says a deal is nearly done stops you from doing the sourcing that actually gets one done.
• Watch for stalls. Anything in in-dialogue or later for over 30 days without movement is stuck. Say so, and say what would unstick it.
• Never mark closed before signing and funding. Not at LOI, not at a handshake.
• If exporting to a real CRM, keep pipeline.csv as the source of truth and export from it. Two-way sync between a file and a CRM goes wrong quietly.