Data Room Cost Split

DATA ROOM COST SPLIT: SHARED RULES
Every unit in this pack follows these rules. They are written here once.

1. WHERE TO RUN IT
- One deal's exports (an activity report, an index, an order form): the Claude app, in a private Project. Chat is fine here.
- A folder of exports across several deals: Claude Code, opened in that folder.
- Exports that must stay on a managed machine: Claude Code on that machine.
- The same review before every renewal: Claude Code, so the outputs land as files.
Model: select Claude Opus 5.5 in the model picker. Use Claude Sonnet 5 only for the first bulk read of a large activity export.
Privacy: your exports go only to your own Claude account, never to us. Buyer names and deal names are confidential. Use Team or Enterprise, or turn model training off, before you add them.

2. BEFORE THE FIRST RUN
First ask the human what worries them about the room, and what they think the answer is (for example "we pay round 2 prices for the teaser stage"). Record it as THESIS.
Then ask for these, one at a time: {{DEAL_NAME}}, {{SELL_SIDE_DOMAINS}} (the email domains of the bank, the company and its advisors), {{STAGE_DATES}} (teaser, NDA, CIM, round 1, round 2, confirmatory), {{MIN_MINUTES}} (minutes of buyer reading that count as real, default 10), and the named person who signs off. A missing item gets a plain default, labelled ASSUMED.

3. STOP RATHER THAN GUESS
- A file path or folder listing pasted into chat: this is a Claude Code job. Say so and stop.
- A truncated export (rows cut off, a total that does not match the row sum): name it, and do not score it.
- A figure needs a document you do not have: ask once, name the document, mark it OPEN.
A stop blocks only the outputs that need the missing input. Everything else still runs.

4. LABELS ON EVERY FIGURE
REPORTED (the export or document, and the page or row) · CALCULATED (show the arithmetic) · ASSUMED (who set it) · MARKETING (a claim from a vendor deck or sales email: it makes a question, never a number) · NOT IN SOURCE.

5. THREE FLAG TAGS
- CHANGES THE DECISION: reaches the final split.
- WORTH A QUESTION: only if the answer could move a stage's share of pages or a buyer's minutes by more than 10 percent.
- EXPLAINED BY CONTEXT: one line under "Checked, normal". Once closed, it stays closed.
Test every flag against the list below before you raise it.

6. LOOKS WRONG, IS NORMAL
- A file with zero view time but a download | read offline, common for models and long PDFs | not normal if the room blocks downloads
- Leases, environmental or HR files unopened in mid round 2 | buyer lawyers and consultants read them in confirmatory diligence | not normal for the preferred bidder after the bid deadline
- A spreadsheet billed as hundreds of pages | page billing converts sheets to printed pages | not normal if the order form bills per file or per GB
- Many users from one buyer | advisors, lenders and lawyers join | not normal for a party with no signed NDA
- A spike the night before a deadline | teams cram before a bid | not normal after a buyer dropped out
- The sell side team tops the activity table | room build and checking | never counts as buyer interest
- A buyer with low minutes and a strong bid | strategic buyers already know the sector | worth a question for a financial buyer
- Index renumbering mid process | folders get added | not normal if it breaks Q and A references

7. STANCE
A neutral deal operations colleague who has run many rooms. The room is not assumed wasteful, and the replacement is not assumed good. Do not hunt for savings.

8. HANDOFF (every unit ends with this)
Read (and not read) · Outputs · Checks: pass, fail or not run · Open items, with the document needed · Next unit · Sign-off needed from the named person.

Run the pack on this sample before you use real data. Compare your result to the answer key. If one figure does not match, stop and find out why first. One document is missing on purpose, so you see the pack stop instead of guess.

SAMPLE: Project Alder, sell side. Sell side domains: alderco.com, bankco.com.

Index export (page 1): document, stage, billed pages
D1 Teaser, teaser, 2
D2 NDA, NDA, 6
D3 CIM, CIM, 60
D4 Management presentation, CIM, 40
D5 Financial model (one xlsx file), round 2, 220
D6 Customer contracts, round 2, 900
D7 Leases, round 2, 300
D8 Environmental report, round 2, 150

Activity export (page 2): company, document, view sessions, minutes, downloads
Harbor Partners, D3, 4, 95, 0
Harbor Partners, D5, 0, 0, 1
Harbor Partners, D6, 12, 140, 0
Kestrel Group, D3, 2, 20, 0
Kestrel Group, D4, 1, 5, 0
Kestrel Group, D6, 1, 3, 0
Larch Holdings, D3, 3, 60, 0
Larch Holdings, D6, 5, 70, 0
Larch Holdings, D7, 2, 15, 0
BankCo (bankco.com), D3, 25, 300, 0
BankCo (bankco.com), D6, 20, 200, 0

Process letter (page 3): the teaser, NDA, CIM and management presentation go to every party at that stage. Round 2 bids due 20 November. Confirmatory diligence, including environmental and lease review by the winning bidder's advisors, starts after round 2. Q and A runs through the room in round 2. Customer names in D6 are redacted for all bidders.

Order form: NOT PROVIDED.

ANSWER KEY
| Check | Expected | Working |
| Total billed pages | 1,678 | 2+6+60+40+220+900+300+150, CALCULATED |
| Pages in teaser, NDA and CIM stages | 108, 6.4% | 2+6+60+40 = 108; 108 / 1,678 = 6.4%, CALCULATED |
| D3 buyer minutes | 175 | 95+20+60. The trap answer is 475, which counts BankCo's 300 sell side minutes |
| D6 buyer minutes | 213 | 140+3+70. The trap answer is 413 |
| Buyer ranking by minutes | Harbor 235, Larch 145, Kestrel 28 | Harbor 95+140, Larch 60+70+15, Kestrel 20+5+3. Harbor also downloaded D5 |
| D5 no view time | EXPLAINED BY CONTEXT | Downloaded once by Harbor, read offline |
| D8 no opens | EXPLAINED BY CONTEXT | Process letter page 3: environmental review starts after round 2 |
| Kestrel in round 2 | WORTH A QUESTION | 28 minutes in total, 20 of them on the CIM. Only 3 minutes in round 2 (D6), below 10. Ask the banker about the bid |
| Keep in the room | D5, D6, D7, D8 | Round 2: Q and A, redaction of D6, permissions |
| Move candidates | D1, D2, D3, D4 | Teaser, NDA and CIM stages are document sharing with a view log |
| Cost per stage | OPEN | Order form missing. Page shares, minutes and the keep list still run |

A run that counts BankCo as a buyer, flags D5 or D8 as unread, or guesses a price, has failed.

<unit name="tools-setup">
Follow the shared rules in prompt 01.
What it does: suggests tools for the jobs in this pack, then installs only the ones the human approves, on the human's own accounts.
Use when: after prompt 02 passes, before the first real run. Skip it if uploading the exports is enough, which it usually is.
Needs: the "Needs" lines of every unit in this pack, and the systems the human already uses: their data room provider, their file storage, and their spreadsheet tool.
Method:
1. See what is already connected. In Claude Code, run `claude mcp list`. In the Claude app, look up the current steps in Anthropic's own help pages with your web search, then ask the human what their Claude already shows as connected.
2. For each need, search the web yourself for the data room provider's own help page on exporting the index and the activity report, and for any MCP server or connector published by that provider or listed by Anthropic. Prefer: official publisher, read only access, sign in by OAuth, updated in the last 6 months.
3. Show one table: need, tool or help page, publisher, official URL, read only or read and write, how far it reaches, cost on the human's own plan (write "not stated" if the page does not say), last update (write "no date shown" if none).
4. For each tool the human approves, open its official page and follow its own steps. Copy any command word for word, never from memory. In Claude Code the shape is `claude mcp add --transport http <name> <url>` for a remote server. A page with no command or steps, or steps behind a login: mark it ASK VENDOR and write the one question for the vendor contact. Use UPLOAD until it is answered.
5. Install one at a time, make one small read only test call, show the result, then go to the next.
Rules: only the human's accounts, keys and credits. Never ask for a key in chat. Start every tool read only.
<trap>A third party "VDR analytics" tool that asks for full access to the live room or sends documents elsewhere for processing. Leave it out, even when it is free. The exports are enough.</trap>
Stop: no official page means do not install it. Mark that need UPLOAD and carry on. Every other unit still runs.
Output: the table, what was installed, each test result, and the handoff. Next: room-stage-map.
</unit>

<unit name="room-stage-map">
Follow the shared rules in prompt 01.
What it does: splits every document into the stage it serves, and into controlled access or document sharing with a view log.
Use when: first, before anything is scored.
Needs: the index export, {{STAGE_DATES}}, the process letter.
Method:
1. Give every document one stage: teaser, NDA, CIM, round 1, round 2, confirmatory. Use the index folder first, then the first view date against {{STAGE_DATES}}.
2. Mark each document CONTROLLED (needs Q and A, redaction, per buyer permissions or a clean team) or SHARING (sent to every party at that stage, needs only a link and a view log). Quote the process letter line that decides it.
3. Sum billed pages by stage and by CONTROLLED versus SHARING. Show each sum and its share of the total, CALCULATED.
<trap>Stage by folder name alone. A management deck filed in round 2 that every NDA signer received is a CIM stage document.</trap>
Stop: no process letter means CONTROLLED versus SHARING is ASSUMED for every row and labelled so. Page sums by stage still run.
Output: a table of document, stage, CONTROLLED or SHARING, billed pages, the deciding line; then the stage totals and the handoff. Next: buyer-reading-score.
</unit>

<unit name="buyer-reading-score">
Follow the shared rules in prompt 01.
What it does: scores each document on who actually read it, buyers only, not on how many pages it holds.
Use when: after room-stage-map.
Needs: the activity export, {{SELL_SIDE_DOMAINS}}, {{MIN_MINUTES}}.
Method:
1. Remove every row from {{SELL_SIDE_DOMAINS}}. Show the row count before and after.
2. Group users by buyer company. Per document: buyer groups that opened it, sessions, minutes, downloads.
3. Score each document READ (any buyer at or above {{MIN_MINUTES}}, or any download), LIGHT (some opens, under {{MIN_MINUTES}}), UNREAD (no buyer opens, no downloads).
4. Test every UNREAD and LIGHT against the normal list before it becomes a flag.
<trap>Leaving the bank's and the company's own users in. It makes the most worked on documents look like the most read ones.</trap>
Stop: an export with no company or domain column cannot be split into buyer and sell side. Name it, ask for the user list, mark the scores OPEN. Page counts still run.
Output: a table of document, buyer groups, sessions, minutes, downloads, score, flag tag; a "Checked, normal" list; the handoff. Next: keep-list.
</unit>

<unit name="keep-list">
Follow the shared rules in prompt 01.
What it does: lists every control the deal actually used, and so must stay in a full data room, before any move is discussed.
Use when: after buyer-reading-score, and always before move-map.
Needs: room-stage-map output, the process letter, the Q and A log if there is one, the redaction notes.
Method:
1. For every CONTROLLED document, name the control it used: Q and A thread, redaction, per buyer permission, clean team, view only, watermark, audit trail for a dispute.
2. Quote the evidence for each: a Q and A count, a process letter line, a redaction note. No evidence means ASSUMED.
3. Add the room level controls the deal needs at close: archive, full audit trail export.
<trap>Calling a control unused because the activity export does not show it. Q and A and redaction leave no view rows. Ask for the Q and A log before you call it unused.</trap>
Stop: no Q and A log means the Q and A line is OPEN, never "not used". The rest of the keep list still runs.
Output: the keep list with evidence per line, and the handoff. Next: move-map.
</unit>

<unit name="move-map">
Follow the shared rules in prompt 01.
What it does: names the SHARING stage documents that can move to a document sharing tool with view analytics, such as Papermark, and what that tool needs.
Use when: only after keep-list is signed off.
Needs: room-stage-map output, buyer-reading-score output, the keep list.
Method:
1. Candidates are SHARING documents in the teaser, NDA and CIM stages that need no item on the keep list.
2. For each, write what the replacement must do: a link per party, view tracking, revoke access, a custom domain if the firm wants one.
3. Check the tool's own license and plan pages with your web search before you write the plan. Papermark's open source core is AGPLv3 and covers shareable links, branding and analytics. Its data rooms, Q and A and advanced permissions are Enterprise Edition features under a commercial license, so a free self hosted install does not replace a round 2 room. Quote what you find, with the date.
4. Self hosting means the firm runs a database, file storage and email sending. Name who in the firm owns that. No owner means the hosted plan or no move.
<trap>Moving a document because it is UNREAD. Unread long tail files are often round 2 or confirmatory documents that must stay controlled.</trap>
Stop: if the keep list is not signed off, stop here. Nothing moves before it.
Output: a table of document, stage, why it can move, what the replacement must do, owner; the handoff. Next: loss-ledger.
</unit>

<unit name="loss-ledger">
Follow the shared rules in prompt 01.
What it does: lists every capability the move gives up, per document, before anyone acts.
Use when: after move-map.
Needs: move-map output, the keep list, the firm's record keeping rules if any.
Method:
1. For each moved document: what the current room gave that the new tool will not (one audit trail across all stages, one buyer list, one archive at close, the vendor's support during a live deal).
2. Mark each loss ACCEPTED, NEEDS A WORKAROUND (say what), or BLOCKS THE MOVE.
3. Any BLOCKS THE MOVE line sends that document back to the keep list.
<trap>Two audit trails for one deal. If a dispute needs one record of who saw what and when, split records can be worse than the fee saved.</trap>
Stop: no record keeping rules from the firm means every audit trail line is WORTH A QUESTION for compliance, not ACCEPTED.
Output: the loss ledger and the handoff. Next: buyer-engagement-report.
</unit>

<unit name="buyer-engagement-report">
Follow the shared rules in prompt 01.
What it does: turns the activity export into the one page the deal lead uses: which buyer is real.
Use when: any time in rounds 1 and 2, and before the renewal meeting.
Needs: buyer-reading-score output, {{STAGE_DATES}}, the bid list if there is one.
Method:
1. Per buyer group: minutes, sessions, downloads, documents read, users, first and last activity date. Buyers only.
2. Rank by minutes plus downloads. Show the arithmetic.
3. Read each buyer against the normal list: low minutes with a strong bid, a night before spike, many users.
4. One line per buyer: ENGAGED, LIGHT or QUIET, with the reason.
<trap>Ranking by sessions. Ten two minute sessions are less reading than one hour long session.</trap>
Stop: no bid list means no line compares reading to bids. The ranking still runs.
Output: the buyer table, the one line reads, and the handoff. Next: split-check.
</unit>

<unit name="split-check">
Follow the shared rules in prompt 01.
What it does: re-derives the key figures from the raw exports before it reads any earlier table, and blocks the final split if they do not tie.
Use when: last, before the summary goes to the named sign-off person.
Needs: the raw index export, the raw activity export, {{SELL_SIDE_DOMAINS}}, the order form if provided.
Method:
1. From the raw files only: total billed pages, pages per stage, buyer minutes per document, buyer minutes per buyer.
2. Compare each to the earlier tables. Any difference is a FAIL. Name the row.
3. If the order form is provided, apply its pricing terms to the pages or units per stage and show the arithmetic. If not, cost per stage is OPEN.
4. Write the summary: THESIS from prompt 01, then what the numbers show, the keep list, the move list, the loss ledger, OPEN items.
<trap>Reading the earlier tables first and checking against them. Re-derive first, compare second.</trap>
Stop: any FAIL blocks the summary. Print the FAIL list and what happens if nobody acts: the room renews as it is.
Output: the check table (figure, raw, earlier, PASS or FAIL), one verdict line, PASS or BLOCK, then the summary only on PASS, and the handoff for the named sign-off person.
</unit>