Build an issue tree for why a tuition centre's enrolment is falling
A tuition centre owner sees fewer new students each term and keeps trying random fixes without knowing the real cause.
An issue tree with at least three levels breaking the falling enrolment into causes, a list of hypotheses ranked by how likely they seem, and the data or conversation that would test each of the top three.
This is a process map project. Draw how work actually flows today, find where it gets stuck, and propose a fix.
The approach: Pick one process with a clear start and end → Map it as it is, not as it should be → Mark delays, rework and hand-offs → Propose one change and what it would save.
Think first
A tuition centre owner sees fewer new students each term and keeps trying random fixes. The question is how to break 'falling enrolment' into separate causes that together cover everything, rank the likely ones, and name a test for the top three before anyone spends money.
- Is the fall in enquiries, in enquiries turning into joins, or in students leaving early?
- How will you make sure no cause sits on two branches at once?
- Which hypotheses are likely given what you know, and which are just possible?
- What one question to a parent or the owner would test each of the top three?
An issue tree at least three levels deep, split first into fewer enquiries, fewer joins and more leavers, with no overlaps, a ranked hypothesis list, and a test for each of the top three hypotheses.
Making a long list of reasons instead of a tree. A list mixes levels and overlaps; a tree splits each branch into parts that do not overlap and together cover the whole.
Gather
- A short conversation with a tuition centre owner or teacher about what has changed, with permission
- Anything the centre already tracks, such as an enquiry register or attendance book, if they share it
- One or two conversations with parents or students about how they chose a centre
AI must not invent enrolment numbers or say which cause is true. Hypotheses stay hypotheses until a real owner, parent or record tests them.
Build with AI
- 3.1 Draft the first level yourselfCLAUDE CHAT
Why this tool: A quick critique of your first split catches overlaps before the tree grows.
Write the three first branches and two sub-branches under each. Ask Claude to find overlaps and gaps.
- 3.2 Grow the tree to three levelsCLAUDE SKILLS
Why this tool: A saved Skill applies the same 'no overlap, no gap' check every time you extend a branch.
Save a short Skill with your rules for a good branch. Run it on each branch as you extend it to the third level, and read every change it proposes.
Before you let it act: Read each suggested fix before accepting it, and keep the tree in your own document so the Skill only suggests changes rather than overwriting your work.
- 3.3 Rank hypotheses and write testsCLAUDE CHAT
Why this tool: Questioning helps you rank by evidence and design tests a small centre can run.
List your hypotheses with what you heard so far, rank them, then ask for help writing a test question for the top three.
Check the AI's work
- The first level splits into enquiries, joins and leavers, or a split you can defend as complete
- No cause appears in two places in the tree
- Each top hypothesis has a test that a real person could answer
- The ranking mentions what you heard, not only what seems likely
- Walk the map with someone who does the work
- Every step has an owner
AI tends to produce a long flat list of causes such as 'online competition' and 'fees too high' with overlapping items, and presents likely-sounding causes as if they were confirmed.
Finish and show
- An issue tree three levels deep
- No overlaps or gaps that you know of
- Hypotheses ranked with reasons
- A test for each of the top three
Draw the tree left to right on one page, highlight the top three hypotheses, and put their tests in a small table underneath.
Show how your first draft of the tree changed after you found an overlap, and which hypothesis moved most after a real conversation.
Which tool fits
- Claude chat: A quick critique of your first split catches overlaps before the tree grows.Free plan works; Pro gives longer conversations.
- Claude Skills: A saved Skill applies the same 'no overlap, no gap' check every time you extend a branch.Pro or above, with code execution switched on.
- Canva (free alternative): drawing the finished mapFree plan is enough for these projects.
- Google Sheets (free alternative): a table of steps, owners and waiting timesFree with a Google account.
Run your three tests with a real centre, update the ranking, and write a one-page note to the owner on what to try first.
AI helps you think, check and polish. The work, and the judgement, stay yours.
This is a free sample. Full Access opens every project in full and lets you submit your finished work.
See Full Access →