# Plan a safe staged switch to a new alert engine

Verrowick · Product-management work sample
All company, customer and data in this exercise are original synthetic material.

## Situation and outcome
Verrowick is a fictional company that sells maintenance and condition-monitoring software to mid-size manufacturing plants, with sensor gateways on the customers' machines. Its alerting pages the on-call technician when a machine reading crosses a threshold. Today that is Classic. Bravquess Routing 2 (Bravquess) replaces it: it adds on-call rosters, escalation chains and grouping of related alerts, because Classic can send one technician nine pages for one fault and runs on a hosted runtime the company must leave. Fourteen sites under ten tenants (customer accounts) depend on it, one of them Verrowick's own test line. The rollout template has four phases: Classic only, side by side (Bravquess watches and logs, Classic pages), Bravquess as the default with Classic as fallback, and Classic removed.

You are the Alerts product manager. You took over the Bravquess rollout plan on Monday 28 September. Today is Monday 5 October. The evidence was collected on Friday 2 October. The rollout review is on Thursday 8 October, where the approvers decide, and the Director of Customer Operations needs tenants and dates by Wednesday 7 October. A release-planning agent drafted a plan on 1 October, and nobody has checked it.

Your outcome is a **proposed staged plan** the approvers can review: for each stage, which tenants and sites, how they are switched, what lets them move on, what makes the team switch back and who does it, what is communicated to whom and when, and which decisions only the approvers can take. A well-argued **blocker record** is equally valid: what is missing, who owns it and when it can be resolved. Running a beta, deciding whether a stage may widen, measuring results, executing Classic's deprecation, running customer migrations, sending notices, approving the plan, switching any engine or fixing the defect are not part of this work. The review itself is not held in this exercise: a note from the release owner is in your download, in the correction note (U001).

The evidence period is 2026-08-10 through 2026-10-02 inclusive. Work with 53 source records (the assignment and decision rights, the rollout template, the contract clause, platform notes, side-by-side comparison exports, customer and field notes and team messages), a table of the sites and an AI-drafted plan. The records are supplied, not run by you. The AI draft is not an authority. Use the supplied source IDs and context; do not contact the fictional teams or seek private data.

## Your role
You are joining the team as Product Manager.

Own the proposed staged plan: tenant order, dates, criteria, triggers, communications and the decisions the approvers must take, or a blocker record if it is not ready. You may propose options with their premises. You may not approve the plan, waive a contract term or a template criterion, change an engine setting, or commit spend such as extended support.

## About the team
A product team replacing its alert engine for customer sites, with a rollout plan reviewed by the SRE manager, the Director of Customer Operations and, for safety-flagged sites, the Safety and Compliance lead.

A release-planning agent drafts rollout plans from the tracker, the rollout channel and the side-by-side dashboard, a QA agent replays differences, and the weekly digest is AI-summarised. People check every claim against its source and own approval, engine settings, customer notices and spend.

Named approvers approve the plan; only the SRE manager changes an engine setting; the Director of Customer Operations owns customer notices and the Safety and Compliance lead signs off safety-flagged sites; extended support is spend that only the approvers commit; the PM proposes and routes but decides none of these.

## Work to complete
1. Read sources/coverage-and-access.md before you make claims about what the evidence covers. The assignment, the decision rights and the rollout template are among the first three records.
2. Audit the AI draft against the original records, row by row. Change only what the evidence warrants. A row that still stands is kept, with its reason.
3. Design the stages. Give each stage its tenants and sites, the engine setting, the default engine and the fallback, measurable entry and exit criteria, the signals you will watch with their thresholds, each rollback trigger as signal, threshold, who acts and how fast, the roster steps, the communications and a checkpoint date. Check every date against what the records allow.
4. Count sites and tenants separately from records: a repeated, copied or re-summarised record is not new evidence. Say what the evidence covers and what it does not, and keep contrary evidence visible.
5. Put the decisions only the approvers can take in front of them, with what each costs and by when, and keep them apart from questions an owner can answer. Never write that something is approved.
6. Read the release owner's note (U001) in your download like any other source and apply it. Carry each change through the plan, the stages, the evidence table and the revision log, and keep the earlier reading visible.
7. Keep to your role. You may propose and route; the people named in the records decide what is theirs.

## Time
You have 90 minutes of work time. The clock starts when you press Start.

After that comes a 10-minute break, then a written interview: up to eight typed questions about your own files, for about 15 minutes. If you submit early, your break starts straight away.

## Mid-task correction
There is no mid-task correction in this take-home. Everything you need is in the files you download at Start, including the correction note **updates/U001-release-owner-note-v1.md**. Read it with the other sources, apply it where it changes your work, and cite it as U001.

## Tools and help
Use any tools and help you like: AI tools and agents, search, your own notes, friends or colleagues. You don't need a paid tool, an API key or an account.

Nothing you do in your own tools is watched. Only your files and your interview answers are reviewed. When you submit, you can say which tools or help you used. That is optional, and leaving it empty is never held against you.

## Submission
Hand in one ZIP with these files at its top level. Start from the copies in the templates folder.
- README.md: list every file you hand in.
- rollout-plan.md: the proposed plan, at most 900 words. Use the template's sections: objectives and sequencing; stage summary; monitoring and rollback; communications and feedback; decisions for the approvers; and, only if the plan is not ready to approve, a blocker record. Write it for approvers who were not in your head.
- stages.csv: one row per stage, at most 25 rows, with the columns of the template. A stage is a group of tenants that move together to one phase on one checkpoint date; the phases are the template's four (Classic only, side by side, Bravquess as the default, Classic removed). Put the phase in the stage name, and when a stage holds several tenants write them in the tenant column separated by semicolons and list all their sites in site_ids. For example: stage "Stage 2, side by side", tenant "Tenant A; Tenant B", site_ids "[their site IDs, separated by semicolons]". Use ISO dates (YYYY-MM-DD) for checkpoint_date and list source IDs in source_ids.
- evidence.csv: your template already has one row for each of the 53 records (R001–R053), with its source ID and site ID filled in. Fill in your own stage name in the stage column, or none, and a disposition such as used, context, repeat, unsupported, contrary, coverage limit, out of scope or changed by correction, with a one-sentence rationale. Leave no record out. Add a row only when a record bears on a second stage; a record that applies to every stage (a fleet-wide rule) gets one row with the stage written as "all", or a list of stage names separated by semicolons. Keep the original site visible and record any correction explicitly. Cite the correction note as U001.
- revision-log.md: material changes to the AI draft, verification checks and remaining gaps, at most 600 words. No private prompts or unrelated files.
- questions.md: at most 10 questions for named owners, each with who should answer and by when.

Keep the six files together under 40,000 characters (about 6,000 words) and each stages.csv cell to a sentence or two: a longer submission cannot be reviewed in full. Words are counted as separated by spaces; table borders (the | characters and the dash rows) are not counted, and headings are not counted.

Use plain text files only: .md, .txt, .csv or .json. Put your work in the files themselves; a link to a document somewhere else does not count. Your ZIP can be up to 5 MB, or 20 MB unzipped, with at most 39 files of up to 2 MB each. Programs, macros, password-protected ZIPs and ZIPs inside your ZIP are not accepted.

Upload your ZIP on your take-home page and wait for **Received and checked**. It normally appears within a few seconds of the upload finishing. If it has not appeared after a minute, upload the ZIP again; if it still does not appear, use **Help and support**. You can replace your ZIP until you submit. Submitting is final and starts your break.

If time runs out before you submit, the last ZIP marked **Received and checked** is handed in for you. If none was received, the team is told that you need support.

If something goes wrong, use **Help and support** on your take-home page. A technical problem is never counted against you. Writing to us does not pause the clock.

## What happens next
The written interview covers these parts of the job:
- Checks what the evidence covers
- Weighs risk and gaps
- Fixes the AI draft
- Hands over a usable plan
- Explains their reasoning
- Designs runnable stages

An AI then drafts a review from your files and your answers. Each point quotes what it rests on. A person at the company reads the review and makes any decision. No one is hired, rejected or ranked automatically.
