Why this is in the vault
A reusable hiring-screen framework — 21 interview questions plus a take-home P&L case study — for evaluating a Controller/finance-ops hire, worth keeping as a template even though RDCO has no open finance role today.
Summary
CJ Gustafson (finance-focused writer behind Mostly Metrics and host of Run the Numbers) publishes his personal interview kit for hiring a Controller, built from his own experience as "Master of Coin" at a startup. The piece is one long list, organized loosely by theme: month-end close process (with a 3-part drill-down on days-to-close, daily milestones, and accuracy checks), cost-saving track record, systems implementation history (including a "tell me about a mistake" question), controls/audit experience, cash governance, order-to-cash and receivables handling, capitalize-vs-expense judgment, overhead allocation, comfort ranking across technical/operational accounting, payroll, tax, and treasury, payment-method judgment (wire/ACH/card), equity plan exposure, and a closing scenario question (new product or country launch).
The 21st and final question is the centerpiece: hand the candidate a real (dummied) chart of accounts / P&L export and ask for 5+ concrete improvements. Gustafson's argument is that a written take-home is the actual signal — it tests Excel modeling, PowerPoint, written communication, and attention to detail simultaneously, and candidates who bail self-select out. He frames the goal as hiring a strategic partner who understands why the business needs certain accounting choices, not a pure number-cruncher.
The newsletter also plugs three "build with Claude + Excel" tutorials (rolling 13-week cash flow forecast, headcount model, net retention cohort model) as a way to stress-test a candidate's ability to "figure it the f out" with AI tooling — relevant as a signal of AI-fluency screening entering finance hiring, not just engineering hiring.
Mapping against Ray Data Co
RDCO has no Controller role and isn't hiring finance ops — this is a case where the specific artifact doesn't map, but the underlying mechanism does. The "one scored take-home over N interview rounds" pattern is the same de-risking logic behind RDCO's own targeting-system prioritization filter (feedback_targeting_system_prioritization_filter) and the cert-escalator path: force a real, gradeable artifact instead of trusting self-reported competence. If RDCO ever brings on a contractor for bookkeeping/finance-ops (a plausible move once cash-flow-tight status per project_income_seat_gap_400k resolves), question 21's live-chart-of-accounts case study is directly reusable — better signal than a generic take-home, since it's graded against the founder's actual account structure. Weak mapping today, portable technique for later.
⚠️ Sponsorship
Two distinct promotional threads, both undisclosed in a formal ad-read sense:
- Rillet (new sponsor, not previously seen from this newsletter) — opens with a "friends of the newsletter" shoutout congratulating Rillet (AI-native ERP) on a fundraise, and states outright that the underlying post was originally written for Rillet's own community/publication ("Close Club"), then republished here. This is closer to co-branded/commissioned content than an arm's-length recommendation — treat the framing (e.g., no explicit pushback on Rillet's product claims) with that in mind. No dollar figure or "sponsored" label disclosed in the email body itself.
- Mostly Talent (self-promo, consistent with known pattern) — CJ links his own recruiting-adjacent services ("I can help you hire" / "get hired") via Typeform. Not a third-party sponsor; same self-consulting flag noted on prior Mostly Metrics notes.
Net effect: the piece's credibility as a neutral hiring guide is intact (the 21 questions themselves aren't Rillet-specific), but the Rillet framing at the top and the AI-tutorial CTAs both carry commercial intent.
Related
- [[2026-08-04-mostlymetrics-cfo-token-cost-gate]]
- [[2026-08-06-mostlymetrics-finance-real-world-report]]