Discovery Template & Question Bank

How to use this

Discovery is structured listening. Each conversation has a target persona, a set of questions calibrated for them, and a defined output you'll add to the synthesis doc. Don't ask all the questions in one meeting — pick the 4–5 that matter most for that conversation, and dig.

Personas in order of typical sequencing:

1. Executive sponsor — the person who signed the contract. Sets vision and definition of success.
2. Customer's data lead — your primary peer. You'll work together for months.
3. Procurement / legal analysts — daily users; their friction is your highest-leverage signal.
4. IT & security — gatekeepers; file tickets early.
5. Output consumers — sometimes a different group from analysts; often higher-leverage stakeholders.

Executive sponsor (30–45 min)

Usually a VP or C-level — CPO, CFO, GC, or a divisional EVP. Don't waste their time on logistics; aim at vision and decision context.

Questions

1. What business outcome are you trying to drive with this deployment?
2. If we delivered exactly the right thing in 90 days, what would change in your operation?
3. What's the metric you'd point a board member at to show this worked?
4. What's the thing you're worried about that I should know early?
5. Who are the three people whose buy-in is most important for this to succeed?
6. Are there freezes or executive priorities I should plan around in the next 90 days?
7. If something has to give — scope, timeline, or quality — which one?

What to capture

• One sentence on the business outcome they want.
• The metric they'd report on.
• The political map (who matters, who's a watch-out).
• Their stated tradeoff preference (the last question is gold; ask it).

Customer's data lead (60–90 min)

Your primary peer. Usually a director or staff-level engineer in their data org. Goal: understand their stack, their constraints, and what they think will be hard.

Questions

1. Walk me through your data stack today — ingestion, warehouse, transformation, BI.
2. Where does the contract / invoice / PO data live today? What format?
3. What's the cleanest path for us to pull data in? What's actually realistic given your team's bandwidth?
4. What's your change-management process for deploying into production?
5. Who else on your team will I work with day-to-day?
6. What's the most painful integration you've done recently, and what made it painful?
7. What's a system in your stack where you'd want to push back if I asked to touch it?
8. Do you have a sandbox / dev environment I can use?
9. What tooling does your team use for orchestration and observability?
10. If I needed to add a new column to a customer-facing dashboard, what's the path?

What to capture

• Stack diagram with system names + which team owns each.
• Their preferred ingestion mechanism (drop bucket, API pull, etc.).
• Change-management process and freeze windows.
• Names of every collaborator you'll touch.
• Their pain points — listen for "the most painful integration" answer; it tells you where to be careful.

Procurement / legal analysts (45–60 min, ideally with 2–3 of them)

The daily users — paralegals, contract analysts, procurement analysts. Their workflow is what the platform will change. Their endorsement is the leading indicator of expansion at renewal.

Questions

1. Walk me through how you'd answer this question today: "what's expiring in the next 90 days that we need to act on?" Every system you touch, every spreadsheet you open.
2. What contracts give you the most trouble? Why?
3. What do you wish you could ask the data but can't today?
4. What part of your workflow is repetitive and you'd hand off to a robot?
5. If you had to pick one number you wish was always accurate and up-to-date, what is it?
6. Where do you keep your own spreadsheets / notes that aren't in any official system?
7. What's a recent contract issue that took longer to resolve than you wanted?

What to capture

• The actual workflow they use today (specific systems, specific friction points).
• The "shadow IT" — spreadsheets and notes that won't show up on the data-stack diagram.
• The repetitive workflow you can replace with the highest leverage.
• The one number they care about most.

IT & security (30–45 min)

Usually one combined call with the customer's IT director + security lead. Lower-frequency contact than the data team, but high-leverage on access.

Questions

1. What's the auth pattern you require for the platform's UI access — SAML SSO, what IdP?
2. Do you require SCIM provisioning for user lifecycle?
3. For service-to-service traffic (our pipeline → your warehouse), what auth do you require?
4. Do you require PrivateLink / VPC peering, or is public-internet OK with TLS?
5. What's your data-residency requirement? Where can our data live?
6. What's the security review process before we can ship to production?
7. Do you have an existing security questionnaire for us to fill out?
8. Who at your end approves new integrations?
9. What's the timeline you typically see for these approvals?

What to capture

• The auth requirements (concrete: which IdP, which protocols).
• The security review process and timeline — the long pole of every deployment.
• Who approves; how to escalate when they're slow.
• Any data-residency or regulatory constraints.

Output consumers (30 min each)

Often different from the analysts. Sometimes it's a VP who saw a demo and signed the contract but doesn't use the system directly; sometimes it's a CFO's chief of staff producing board materials. They define what success looks like in artifact form.

Questions

1. How would you use this output in your day-to-day or monthly work?
2. What format do you need the output in — dashboard, weekly email, embedded in another tool?
3. What's a decision you've recently made where this data would have changed your call?
4. If we showed you a dashboard at the end of month one, what charts would you most want to see?
5. Who shares this output with you today, and how?

What to capture

• The artifact format they need (dashboard, report, etc.).
• Specific charts / numbers they'd want to see.
• Existing reporting they expect to replace or augment.

Reading the corpus (10–15 hours over the first two weeks)

Read at least 20 representative documents from the customer's corpus. Not for completeness — for pattern. What you're looking for:

• Layout variance. Are all the MSAs visually similar, or does each year / template version look different? Are they templated or fully bespoke?
• Completeness. Are amendments present? Are SOWs separately filed? Do you see signature pages, or do they trail off?
• Age range. Pre-2010 contracts may use different vocabulary and weaker scans. The full corpus often spans 15+ years.
• Language and jurisdiction. Multi-region customers have English / French / German / Mandarin contracts. Plan for translation if needed.
• Quality of scans. Native PDF, image-based, mixed, handwritten amendments?
• Edge cases. What weird things show up — embedded contracts inside emails? Multi-tenant contracts where one signed doc covers multiple buying entities? Bundled SOWs?

Take notes per document class. By end of week two, you should have a clear sense of where extraction will work well and where it will struggle.

Synthesis template

Use this Markdown template to consolidate discovery into a draft before writing the working agreement. Share with the customer's data lead for accuracy check before going to the sponsor.

# [Customer] — Discovery Synthesis
Last updated: YYYY-MM-DD

## Business outcome
- Stated outcome:
- Metric the sponsor would point to:
- Confidence we can deliver this in 90 days:

## Stakeholders
- Executive sponsor: [name, role]
- Data lead (primary partner): [name, role]
- Procurement / legal analysts: [names]
- IT / security gatekeeper: [name]
- Output consumers: [names + what they want]

## Stack
- Source documents: [where they live, what format]
- CLM: [if any]
- ERP / system of record: [vendor + version]
- Procurement suite: [if any]
- Warehouse: [vendor]
- BI: [vendor]
- Identity: [vendor]
- Orchestration / observability: [vendors]

## Corpus profile (from reading 20+ docs)
- Document classes in scope, ranked by importance:
  1.
  2.
  3.
- Layout variance per class: [low / medium / high, notes]
- Completeness: [amendments present? signatures? SOWs linked?]
- Age range:
- Languages / jurisdictions:
- Scan quality: [native / image / mixed; specific concerns]
- Edge cases observed:

## Workflow being replaced
- Current manual process:
- Hours per week today:
- Pain points called out:

## Constraints
- Change-management process:
- Freeze windows:
- Security review timeline:
- Data residency requirement:
- Other regulatory:

## Risks (ranked)
1. [Risk]: [mitigation]
2.
3.

## Proposed first slice
- Document class:
- Source:
- Output target:
- Success criterion:

## Open questions
-
-
-

This synthesis doc becomes the input for the working agreement (next chapter).

✓ Check yourself

Answer from memory before opening each one.

1. You have 45 minutes with the executive sponsor and seven prepared questions. Which answer do you most need to leave with, and why does no other interview give it to you?

   Show answer

   Their stated trade-off, which of scope, timeline or quality gives first. Nobody else can tell you, and you will need it the first time you are forced to choose. The outcome, the metric and the political map are the other three things to capture. See the sponsor conversation.

2. A procurement analyst answers your first question with a long walk through five spreadsheets and two systems. Why is that the best use of the hour, and what do you listen for?

   Show answer

   The question asks them to answer a real recurring question today, step by step, so it exposes the actual workflow. Listen for shadow systems that never appear on a stack diagram, the sources they distrust and the number they wish was always right. Letting them ramble is the point. See the analyst conversation.

3. Your stack diagram is still half blank in week one. Why do you ask the IT and security questions anyway, and which answers matter most?

   Show answer

   Security review and access approvals are the long pole of every deployment, so you need the timeline and the approver early. The concrete answers are the identity provider and protocol, whether SCIM is required, how service traffic authenticates, network path and data residency. See the IT and security call.

4. You read 20 documents the customer chose for you. All are clean native PDFs from the last three years. What do you conclude, and what do you ask for next?

   Show answer

   Treat the sample as the best end of the corpus, not a picture of it. Corpora often span 15 or more years with weaker scans and different vocabulary. Ask for a sample spread across years, sources and scan quality, and run the OCR preflight on it. See reading the corpus.

5. Why does the data lead see the synthesis document before the sponsor does?

   Show answer

   The data lead is your peer and can correct facts about the stack and warn you about political problems cheaply and privately. A sponsor who reads a document with wrong facts loses confidence in the work before it begins. See the synthesis template.

6. The sponsor wants a board-level summary and the analysts want a drill-down dashboard. Both say they are the main consumer. How does discovery handle it?

   Show answer

   Capture both artifact formats under output consumers, ask the sponsor for the trade-off they prefer, and pick one first-slice output with a stated success criterion. If the two groups still disagree, note it as a ranked risk in the synthesis rather than building both. See output consumers.

Takeaway

• Discovery is structured listening: for each person, pick the four or five questions that matter and dig, instead of running the whole list.
• From the sponsor you need one sentence on the outcome, the metric they would report, the political map and their preferred trade-off.
• The data lead gives you the stack, the preferred ingestion route, the change process and the places they would push back.
• Ask analysts to walk through how they answer a real question today; it exposes shadow systems and the number they care about.
• Start IT and security early, because access and review time sets the schedule more than the build does.
• Reading 20 or more documents for pattern shows layout variance, completeness, age, language, scan quality and edge cases.
• The synthesis document consolidates all of it, and the data lead checks it before it goes anywhere else.

The synthesis document is the input to the one-page agreement you will ask the sponsor to sign. → Working Agreement Template

Further reading

• Dimensional Modeling Techniques — Kimball Group, accessed 2026. Business-process and grain thinking for requirements gathering; a frame for what to ask the analysts and consumers.
• Overview of federated authentication and SSO — Snowflake documentation, accessed 2026. The SSO and identity-provider vocabulary behind the IT call, using one warehouse as the example.
• Snowflake SCIM support — Snowflake documentation, accessed 2026. What SCIM provisioning covers, so you can judge whether the customer's requirement is realistic.