Backsolved opens to its first users in September 2026. See pricing and early-access seats →

Backsolved

Case interview practice

Your framework was fine. Your arithmetic wasn’t.

Backsolved runs a full consulting-style case interview, then checks every number you said against the case’s fixed ground truth — line by line, before it grades anything else.

First case free, including the full report. No card to start.

Arithmetic check · extract from a real report
“revenue is flat at 2.4bn, so 2.4bn − 1.9bn = 500 million” checks out
“the margin is (500 − 450) / 500 = 1.0%” off by 100× Likely a percentage-as-fraction mix-up. The margin is 10%, not 1.0%. You carried 1.0% into the recommendation two turns later.
“loan losses went up by about $50M” within tolerance Ground truth for this case: $50M. You cited it on turn 9.
“so default rates roughly doubled — call it 9%” off target Ground truth: 7.4%. Directionally right, numerically loose — and you stated it with more confidence than the working supported.

Illustrative extract, using the ground-truth values from the free retail-banking case.

You cannot grade your own arithmetic

That is the actual problem with practising cases alone. You can read a framework guide and know whether you covered the right buckets. You can time yourself. What you cannot do is notice, in the moment, that you divided when you meant to multiply — because if you had noticed, you would not have done it.

Practice partners miss it too. They are following your logic, not auditing your figures, and half the time they do not know the right answer either. A general-purpose chatbot is worse than useless here: it will confidently agree with your wrong number, then build the rest of the case on top of it.

So the mistake survives. It survives practice, it survives the mock, and it turns up in the round that counted.

How it works

01

You take a real case

A written interview with a specific interviewer, a specific company and a specific problem. You ask for the data you want; you only get what you actually ask for. Roughly 30–40 minutes, typed, at whatever hour suits you.

02

Every number is checked

Each case carries a fixed answer key written before you ever saw it — the real drivers, the real figures, the tolerance each one is allowed. Your equations are recomputed. Your citations are compared against the key.

03

You get the whole report

Scores on four dimensions with the quotes behind them, framework coverage, the full arithmetic table, the ground truth revealed, and the transcript annotated turn by turn. Free case included.

Why this isn’t another chatbot with a prompt

The distinction is not the model. It is where the answer lives.

  A general AI tool Backsolved
Where the answer comes from Generated on the spot, alongside your reasoning. Fixed before the case starts and stored where you cannot read it.
Who does the arithmetic The model — which is why it sometimes agrees with maths that does not compute. Ordinary code. Your equations are literally recomputed and compared.
If you state a wrong number It is usually accepted and built upon. It is flagged, with the correct value and the size of the error.
Consistency between two runs The “right answer” can move between attempts. The key is the same every time, for every candidate.
Asking for data you didn’t earn Tends to hand you everything at once. Facts are released only when you actually ask the question.

The interviewer’s conversation is written by a language model. The scoring of your figures is not — that part is arithmetic, and it is the part we are willing to be held to.

Who this is for

  • MBA candidates preparing for consulting first rounds.
  • Undergraduates and postgraduates going through campus placement season.
  • Working professionals moving into strategy, consulting or corporate development.
  • Anyone facing a case-style interview — the format now shows up well beyond consulting.

It suits people who already know what a profitability tree is and want the reps, more than people meeting the case format for the first time. If you are starting from zero, read a framework guide first, then come back.

Pricing

One payment. 120 days. No subscription, no auto-renewal, nothing to cancel.

First case

Free

Including the complete report.

  • A full 30–40 minute case
  • The entire report, not a teaser
  • Arithmetic check and ground-truth reveal
  • No card required

The first 50 people pay ₹799 / $15. That is the whole early-access offer — no other terms attached to it.

Questions people ask first

Is this available right now?

Backsolved opens to its first users in September 2026. The product is built and in final testing; what remains is the payment and account setup. Email support@backsolved.com and we will tell you the day the free case opens.

Does it grade me the way a real interviewer would?

No, and we will not claim it does. It measures things that can be measured honestly: whether your arithmetic is right, whether you covered the case’s drivers, whether you stated a structure before reaching for numbers, how you paced yourself. The written judgment on top of that is generated automatically and should be read as practice feedback, not as an assessment of how any firm would rate you.

Is my transcript private?

Your cases and reports are yours. They are not sold, not shared, and not used to train anybody’s model. You can delete your account and everything in it at any time. The privacy policy spells out exactly what is stored and where.

What if it doesn’t help?

Take the free case first — the whole report, not a sample. If you do buy and it is not what you needed, email us within 7 days and we refund it. See the refund policy.