A voice interviewer isn’t enough.
The opportunity to test is clearer, faster feedback that changes the next practice session. We have not established that competitors lack this inside their paid products.
What was actually tested
| Product | Observed strengths | Access / limits |
|---|---|---|
| RocketBlocks | Case library with industry/type/difficulty filters. Marathon Monday offers typed mock, AI voice mock, PDF and expert video together. | Both mock buttons triggered an account/subscription gate. Signup shows $35/month, a 7-day trial and password creation. No mock or feedback completed. |
| CaseCoach | Structured toolkit, preparation plan, courses, cases, drills and peer practice. The free trial explicitly lists a limited resource set. | 10-day free trial registration requires email, name, country and password. Awaiting user-completed signup. No drill scoring or peer session tested. Toolkit advertised $49/month or $199 annual access. |
| PrepLounge | Detailed case library, employer cases, peer slots, coach/community answers. Opened BeautyCo and successfully revealed its short solution without login. | Public case flow is reading/revealing, with no observed answer capture in that flow. Google signup approved and advanced to profile setup; extra profile details not supplied. Peer session and personalized feedback untested. |
| IB Vine | Earlier public sample test: compact question card, answer reveal and a detailed accounting walkthrough. A useful reference for a direct question-to-answer experience. | Different subject matter. Public sample tested, not a full paid voice mock. Placement cards and marketing usage counts are not causal evidence of effectiveness. |
Prices are displayed offers observed during this review, not purchases. Subscription terms, taxes and available content may vary. Library and signup case counts differ on RocketBlocks; do not interpret either as independently audited.
What we could make noticeably better
1. First useful practice before signup. Start one original case immediately. The demo does this; whether it improves activation must be measured.
2. Explain the exact error. Show the submitted number, reference calculation and business implication. Avoid a vague “7/10” or an ungrounded hiring-readiness score.
3. Close the learning loop. Retry the same skill with new numbers and preserve the first attempt for a reviewer. This demo uses a fixed retry; adaptive remediation would require more exercises and validation.
4. Make feedback trustworthy. Ground arithmetic in an explicit case model and have experienced interviewers review written reasoning. The demo deliberately does not pretend to grade prose with AI.
What this demo includes
The consulting track includes one fictional bike-business case: structure, exhibit math, a challenging expansion decision, six checked calculations, a changed-number retry and a downloadable attempt. No account, backend, voice model or payment collection. Written work is held in memory and cleared on reload/close.
Learning feedback upgrade
Added separate calculation and explicit concept checks across16 scenarios, targeted numerical error explanations, and first-attempt versus changed-number retry history. The voice interviewer can call a deterministic numerical checker. Typed prose is not automatically semantically graded. The browser test verified a correct UFCF number paired with a wrong financing explanation is flagged, and the subsequent retry preserves both attempts. This is evidence of implementation behavior, not competitive superiority or improved real interview outcomes.
New IB track
The IB prototype adds an original three-statement exercise, a valuation bridge, one of three numerical-answer-based follow-ups, worked feedback and a transfer exercise. It is typed, not voice-driven. Its effectiveness relative to IB Vine is unproven. Candidate invitations and a five-person test protocol are prepared as drafts; nobody has been recruited yet.
Candidate test before a platform
Recruit 5–8 consulting candidates with interviews in the next 4–8 weeks. Keep outreach as drafts. Have each try our case plus a comparable competitor task they can access; counterbalance order and match difficulty. Ask permission before recording anything.
Measure time to first attempt, completion, explanation of their own error, an unseen follow-up problem, and which experience they would use again. Have an experienced consultant score reasoning blind to product. Satisfaction alone does not prove better interview performance.
Payment test: after a successful free session, offer a clearly scoped, human-reviewed three-case pilot at a proposed $49. This is a price hypothesis, not a market benchmark. Count completed purchases only once delivery and checkout exist. There is no checkout here and no paid validation yet. At 45 minutes of review at $60/hour, $49 leaves almost nothing after payment fees; price, labor or delivery scope must change before scaling.
Decision: continue only if candidates prefer the correction/retry loop and some genuinely pay. A prettier library or another general AI interviewer is not a defensible business. Longer-term advantage would need expert-calibrated evaluations, measurable learning gains and low-cost access through clubs or schools.
IBMock public report inspection · 19 October 2026
A public basic-accounting session provides question-level commentary, numerical walkthroughs, strengths, focus areas and recommended next steps. This is materially richer than a generic score. We inspected a public report, not a live interview, and did not test paid sessions or verify score calibration.
One working-capital item contains inconsistent guidance: feedback penalizes inclusion of cash and current debt, while its example answer uses a broad working-capital definition including cash. This supports testing definition consistency; it does not establish that Accretive Thinking performs better overall. Our accounting guide now explicitly distinguishes broad and operating working capital.
Public report inspected ↗ · Our working-capital explanation →