The product
Practise the job,
don’t just read about it.
A course tells you what good looks like. RoleCube gives you a role, a few working days of conflicting messages, meetings and data, and a decision that’s yours to make. Then it shows you how a manager would have read it.
01 · How it works
One scenario, several working days.
Every scenario is written end to end: the company, the people, the data and the answer key. What changes from one attempt to the next is you: what you read, who you ask and when you decide you know enough.
- 01
Choose a role and a level
Each scenario is written for a specific role, such as Product Manager or Data Analyst, at a specific seniority. Before you start, it lists the skills it covers and how its days are planned, so you can pick the practice you need.
- 02
Get briefed like a new starter
A short briefing gives you the company, your place in it, the people you’ll work with and the assignment. It’s the context a real first week would give you before anyone expects an answer. You can skip it if you have played before.
- 03
Work the day as it arrives
Each day opens with objectives and a clock that runs from 08:00 to 18:00. Messages, meetings and documents are released day by day, and nothing from a later day can be read early. You’re always deciding with what you have, not with hindsight.
- 04
Decide what the evidence is worth
The information is incomplete on purpose. Two stakeholders hold opposing views and both are partly right, the data supports more than one conclusion, and a document quietly contradicts a message. You won’t get through everything, so choosing what to chase is part of the skill.
- 05
End the day on your own terms
At six o’clock you choose: close the day with what you have, or stay on in overtime until midnight. Which threads you followed and which you left is part of the record, so you feel the trade-off between thoroughness and time.
- 06
Make the call
On the final day you write the report: the root cause you believe, how confident you are, the priority and owner, the queries that back it up, what should happen next and who needs to read it. Then you submit it. There’s no second opinion and no answer at the back.
02 · The tools
Four tools. Each builds a different skill.
MessagesTrains stakeholder management
Every thread is a person with a point of view.
Stakeholders write to you in character and reply to what you actually say, based on the scenario’s own script. You practise something that’s rarely taught: asking the right person the right question early enough for the answer to be useful.
- Replies are based on what that stakeholder knows, not improvised
- Who you brief before you submit is part of the evaluation
- Pressure arrives on its own schedule, like it does at work
Sofia Okafor · VP Product
Messages
MeetingsTrains structured meetings
Sit in the room, then ask what you missed.
Meetings play out as a transcript of people disagreeing in real time, and afterwards you can put your own questions to the room. You learn to walk in knowing what you need to leave with, and to spot a contradiction while it’s still in front of you.
- Attendance is recorded, so skipping the steering meeting is a choice
- Follow-up questions are answered from the meeting’s own material
- Conflicting accounts are left for you to reconcile
Steering meeting
Live · 3 present
- 14:06
Amir MalikEngineering
The release is stable. Error rates didn’t move.
- 14:07
Lena HaddadCustomer Success
I’m not saying it broke. I’m saying nobody knew what changed.
- 14:09
Sofia OkaforVP Product
Then someone tell me which of those we’re fixing this quarter.
Your turn to speak. What you say changes how they respond.
Data workspaceTrains data literacy
Query the tables and know when to stop.
Each scenario comes with its own tables, which you query yourself in a read-only workspace. Any query can be attached to your report as evidence, so the numbers you cite are numbers you actually pulled.
- Scenario tables, queried read-only
- Query history is kept, and attached queries travel with the report
- The headline figure is rarely the one that matters
customer_retention
Warehouse
| Segment | Accounts | Churn |
|---|---|---|
| Enterprise | 184 | 3.2% |
| Mid-market | 426 | 6.7% |
| SMB annual | 1,208 | 12.9% |
select segment, accounts, churn_rate
from customer_retention
order by churn_rate descThe headline number isn’t the one the board is worried about. Working that out is the exercise.
DocumentsTrains synthesis
Twelve notes and no summary.
Cancellation notes, specs, board packs and briefs arrive unsorted and unhighlighted. You take away the skill of turning a pile of partial accounts into one clear argument, and showing where it came from.
- Which documents you opened is part of the record
- Nothing is pre-highlighted for you
- Evidence in the report links back to its source
Cancellation notes · Q3
Document
Account 0412 · Mid-market · Cancelled 14 Aug
“The product does what it says. Our problem was the first ninety days. We were handed from the sales team to an onboarding manager to a support queue, and each of them asked us the same questions again.”
“By the time anyone understood our setup, the team had stopped logging in.”
“Renewal was a formality we no longer had a reason to complete. Nobody internally could name what we were paying for.”
Note 3 of 12 · Exactly as it arrived
03 · The evaluation
Graded against a rubric.
When you submit, you get the review a good manager would write, specific to the decisions you made and the evidence you used. Underneath it’s a rubric written with the scenario, not a number made up by a model.
What you get back
- A score and grade, with the result for every rubric criterion
- Which outcomes you met, partly met or missed
- A manager-style review, in plain language, of the decisions you made
- A replay, if you want to run the same week differently
Want to know what the AI decides, what happens if it’s unavailable, or whether hints count against you? See the grading questions in the FAQ.
How a score is built
Evaluation
Process
Computed in code
What you did: stakeholders briefed, meetings attended, documents opened and evidence attached. Hints, overtime and lateness are noted in your feedback but never reduce your score.
Rubric
Each criterion 0–4
Every criterion is assessed against the scenario’s answer key, with AI assistance, and your own writing has to support the root cause you picked. A lucky guess doesn’t keep full credit.
Blend
Weights set per scenario
The two are combined using weights set for that scenario. The arithmetic runs in code, so the final number can always be traced back.
Why you can trust the number
The AI judges criteria. It never decides the facts of the scenario, the results of your queries or the final arithmetic. If it’s unavailable, the deterministic score stands and the review is flagged for a human, so an outage never costs you points. Your evaluation is saved with the attempt, so it reads the same every time you open it.
04 · What you take away
What you keep after the scenario.
- 01
A call you have defended
You have committed to a recommendation with incomplete evidence, under a deadline, and heard it assessed. You can’t get that practice from reading.
- 02
Evidence habits that hold up
Citing the query you ran and the document you read, rather than the number you remember, is a habit you build here and keep.
- 03
A record you can compare
Every attempt keeps its report, score and review, and you can replay a scenario to run the same week differently. That lets you see how you’re improving over time.