14 min · Advanced
Case Study: Northline Coffee Roasters
Work a fictional six-person roastery through a modest Gemini strategy: two office segments, one vague ad killed, and no miracle ending.
Learning objectives
- Separate Northline's notes into facts, hypotheses, and missing sources.
- Run a problem prompt, a segment prompt, and a strategy prompt without inventing a win.
- Name what the team learned, including the tally that still has to happen outside the chat.
01
Why this matters
Case work is how you practice judgment before you spend money. Northline is fictional, and the point is the method, not a fairy-tale win. Killing a vague ad is a real outcome. Leaving a rainy-week tally blank is also a real outcome.
02
Concept
Northline Coffee Roasters is a fictional six-person roastery. They sell beans to offices and run a Saturday stall. Offices are pausing subscriptions. The stall is busy, yet rainy weeks lose money because staff and prep assume sun. This lesson does not end in a miracle. You will not be told that a prompt saved the company. You will see a modest use of Gemini: separate facts from guesses and kill a vague idea.
The Gemini strategy reads the notes. It does not invent a brand. Keep office buyers and stall buyers apart. Ask which office segment still has a reason to reorder, and which stall costs move when rain cuts the crowd. Refuse any plan that needs a figure the notes do not contain. The strategy prompt in the breakdown demands two segments, one thing to stop, and the facts still missing.
You also need two partner prompts beside that strategy prompt. The problem prompt forces every statement into fact from the notes, hypothesis, or needs a source. The segment prompt takes only the office notes and returns two groups, each with a job-to-be-done, a quote you paste, and a reason the group might say no. Neither prompt may invent a win. If the notes are thin, the right output is a short list of questions for the next real conversation.
The process is sequential. Paste notes, run the problem prompt, run the segment prompt, run the strategy prompt, then delete anything cute. The modest result: the team named two office segments, a steady kitchen account and a paused reception account, and killed a vague community ad that could not say who it was for. Lessons: segments need evidence, a killed idea counts, and rainy-week costs still need a tally outside the chat.
03
Real-world example
Three prompts, no rescue ending
Three prompts carry this case, and none of them may invent a win. Problem prompt: Restate Northline using only the notes I paste. Label each line fact from the notes, hypothesis, or needs a source. Do not add counts or a happy ending. Segment prompt: Using office notes only, name two segments. For each, give the job-to-be-done, a quote I pasted, or missing quote, and a reason they might say no. Strategy prompt: Ask for those two segments, one rainy-week note, and one item to stop, with those labels and the headings Segments, Stop, and Missing. Run them in order. Then cut any victory line.
04
Gemini demonstration
A realistic exchange. Open Gemini beside this page and try a version with your own facts.
You
Problem prompt. Role: analyst who may not add facts. Notes: Northline Coffee Roasters, six people, beans for offices, Saturday stall. Some offices paused subscriptions. Stall can be busy. Rainy weeks still pay staff and prep. Task: restate the problem. Label every sentence fact from the notes, hypothesis, or needs a source. Do not invent paused counts, losses, quotes, or a happy ending.
Gemini
Problem sort, using only the case notes you pasted. Fact from the notes: Northline is a six-person roastery, sells beans to offices, runs a Saturday stall, and some offices have paused subscriptions. Fact from the notes: rainy weeks are a problem because staff and prep still happen. Needs a source: any exact loss, any count of paused offices, any claim that the stall is unprofitable by a stated amount. Hypothesis, if you want it labeled: reception accounts may pause faster than kitchen accounts. I will not turn that into a result. I will not add a saved-revenue ending. The next prompt should not ask for a brand campaign. It should ask which statements are still guesses. Outside the chat, someone still has to tally a rainy Saturday.
You
Segment prompt. Use only office notes. My notes: one kitchen account has not paused. Two reception accounts paused. I have no quotes pasted. Return exactly two office segments. For each give the job-to-be-done, missing quote if I gave none, and a reason they might say no. Do not add a community segment. Do not invent a win.
Gemini
Two office segments, and no third segment called everyone. Kitchen accounts: job-to-be-done is a steady bean supply for people who drink at work. Evidence: your note that one kitchen has not paused. Reason they might say no: price or a change in headcount, which you have not measured, so that reason is a hypothesis. Reception accounts: job-to-be-done was a lobby gesture, and your note says two of these paused. Reason they might say no: the display no longer feels worth the subscription. I did not invent quotes. Where you did not paste words, the cell says missing quote. The vague community ad should stop because it names no segment and no job. Stall rain costs remain needs a source. This is not a turnaround story.
05
Prompt breakdown
The prompt
You are a careful analyst for a six-person roastery. Use only the notes I paste about Northline Coffee Roasters. Do not invent sales, savings, quotes, or a rescue ending. Task: write a strategy with exactly two office segments, one stall note on rainy weeks, and one item to stop. For each segment state the job-to-be-done, the evidence from my notes, and a reason they might say no. Label every line fact from me, hypothesis, or needs a source. Output headings: Segments, Stop, Missing. If evidence is thin, list what I must measure outside this chat, including a rainy Saturday tally. Do not recommend a community ad that names no one.
- The opening limits the model to pasted notes, which is the only evidence this case allows.
- Two segments and one stop force a choice instead of a vague community message.
- Labels on every line keep a hypothesis from dressing up as a result.
- The Missing heading, including the rainy tally, blocks a miracle ending.
06
Common mistakes
Usually
Write the comeback story and include new sales figures that sound realistic.
Better
Use only my notes. If a figure is missing, label it needs a source.
Usually
Recommend a community ad so the brand feels human.
Better
Stop the community ad if it cannot name a segment and a job.
Usually
Invent quotes from office managers about how much they love the beans.
Better
Write missing quote when I did not paste words. Do not fabricate a testimonial.
07
Practice exercise
Run the three prompts in order
Done when: You have two office segments, one stopped idea, and a missing rainy-week number that you did not invent.
08
Challenge
What you would stop, and what you would count
Using only the Northline notes in this lesson, list two office segments and one thing you would stop. For each segment give the job-to-be-done, the evidence you actually have, and a reason they might say no. Mark rainy-week losses as needs a source unless you invent nothing and instead state what to measure on a wet Saturday. Do not write a rescue ending or a community campaign. You are done when a teammate can see the killed idea and the missing tally side by side.
Hint: The kitchen account and the reception account are different jobs. If your paragraph treats every office as the same buyer, split them.
09
Key takeaways
- Northline is a practice case: offices pausing, a stall that struggles in the rain, no miracle.
- The process is problem prompt, segment prompt, then the strategy prompt.
- The modest result was two office segments and a killed community ad.
- A killed vague idea counts. A rainy-week tally still has to be done outside the chat.
10