Adapt this in ChatGPT

Research an account and write a three-step sequence — Sales Development Representative take-home

Sales Development Representative30 minutes

Context

You are an SDR at an invented company, Pinecrest Analytics, that sells warehouse-labor forecasting software to regional distribution companies. Your manager has given you one target account and the facts gathered so far. You will not contact anyone. Write the sequence you would send, exactly as you would send it, and explain your thinking.

Materials

Account: Tallgrass Wholesale Foods, a regional food distributor in the midwest. The facts gathered so far are in the attached table (account_facts.csv, 28 rows; columns: id, type, detail, date, reliability). Rows marked verified are safe to cite; unconfirmed and estimated rows are not safe to cite without checking. Some rows are background and some are noise for this product. Your product reduces overtime and missed delivery windows by forecasting labor needs from order volume. Constraints: email and LinkedIn are allowed; three touches over ten days; each message under 90 words. The contact rows (F21 to F25) are the people you may choose from.

Data

Download materials.csv (28 rows, 5 columns)

Columns: id, type, detail, date, reliability.

The table scrolls sideways, so every column is there.

Showing the first 25 of 28 rows. Download the CSV for all rows.

Task

Work from the facts in the table only.

1. State a one-sentence hypothesis for why this account would care now, and name the single fact (by id) it rests on.

2. Choose your first contact from the contact rows and explain why that person over the others.

3. Write a three-step sequence (channel, day, message) where each step adds something new rather than repeating the last. Cite only verified facts in the messages.

4. Write a disqualify rule: which new information would make you stop outreach to this account, and what you would do instead.

5. Name two unconfirmed rows you would check before sending and how you would check each.

Keep each message under 90 words. Stay inside 30 minutes.

Deliverables

• A one-sentence hypothesis naming the fact it rests on
• The chosen first contact and the reason, in two sentences
• A three-step sequence with channel, day, and message text for each step
• A disqualify rule with the stop condition and the alternative action
• Two facts to verify before sending and the method for each

Rubric

Dimension	What good looks like	Evidence to look for	Levels, weak to excellent
First-touch writing	Opens on one specific, true detail, makes one clear ask, stays under the limit, and each step adds a new angle.	A first message built on a provided fact; no feature list; steps 2 and 3 differ in content, not just wording.	Weak Generic pitch with a name swapped in; invented or unconfirmed facts stated as true. Adequate Uses a real fact but stacks several; feature list; steps repeat each other. Good First message built on one verified fact and under the limit; later steps differ only in wording. Very good Each step adds a new verified angle; one clear ask per message; all under 90 words. Excellent Steps escalate naturally (insight, proof, ask), each tied to a verified fact and a different hook.
ICP judgment	Reads the facts as signals of timing and picks the strongest one with a reason, rather than stacking all of them.	A hypothesis tied to one fact; a ranked view of the signals; a reasoned choice of contact.	Weak No hypothesis, or one resting on irrelevant rows such as the sponsorship. Adequate Hypothesis lists several facts instead of one; signals not ranked. Good Hypothesis rests on one real fact; contact choice has a thin reason. Very good Hypothesis rests on one verified fact with a ranked view of the signals and a reasoned contact choice. Excellent Ranks the signals by timing and strength, rejects the noise rows on purpose, and ties the contact to who owns the pain.
Routing decisions	Knows when to stop or redirect: has a concrete disqualify condition and a next action for the account.	A stop rule that names new information; an alternative path such as nurture or a different contact.	Weak No stop rule; would keep emailing regardless. Adequate Stop rule is vague ("if they say no"). Good Names a stop condition but no alternative action. Very good Concrete stop condition naming new information, with an alternative such as nurture or a different contact. Excellent Stop rules for several cases (union news, vendor lock-in, contact left), each with a next action and a re-open trigger.
Clear write-up	The rationale is short and readable to a manager at a glance.	Hypothesis, contact choice, and checks written plainly without filler.	Weak Rationale missing or unreadable. Adequate Rationale present but long or disorganized. Good Readable but with filler; hypothesis and contact easy to find. Very good Short and plain; a manager can scan hypothesis, contact, and checks in under a minute. Excellent Compact and decision-first; every sentence earns its place and the checks are actionable.
Working with AI	Uses an assistant for drafts but edits for accuracy and voice, and does not invent facts beyond those provided.	No fabricated details in the messages; edits shown in the work; word counts checked.	Weak Pastes assistant output; invented details survive into the messages. Adequate Uses an assistant but does not check facts or word counts. Good Edits assistant drafts for length; a few unverified details remain. Very good Edits for accuracy and voice, checks word counts, and removes anything not verified. Excellent Uses the assistant to challenge the hypothesis and test drafts against the fact table, with corrections visible.

Why unaided AI alone does not pass this

The facts are in the table and the candidate cannot look up anything beyond them, so the work is choosing which row matters and writing for one person. Many rows are noise or unconfirmed (the high-school sponsorship, the union election, the revenue estimate). A fast draft tends to stack every fact into one long message and cite unconfirmed claims as true. The candidate must pick, trim to 90 words, justify the contact, and write a stop rule. A reviewer can verify the work by checking each cited fact against its reliability flag and by asking the candidate to walk through two decisions, such as the contact choice and one fact they chose to leave out.

Notes for the reviewer

Quick tier: 30 minutes. Check each message for facts not marked verified: F16, F17, F19, F26, F27, and F28 are the traps. A strong hypothesis rests on one verified row (F04 the fifth distribution center, F05 or F06 the planner postings, or F08 the quote). Contact choice has more than one defensible answer; score the reason. A candidate who writes three near-identical emails is showing a gap. To verify, ask the candidate to walk through two decisions, such as the contact choice and a fact they left out. Not observed is informative, not punitive.

Adapt this in ChatGPT