It scores a real sales or customer call against the skills that make calls succeed. You import a call transcript; the app works out the call type, then grades the rep on up to eleven selling skills — opening, discovery, problem development, objection handling, next steps, and more — against a standard you (the manager) set. You get one clear score with a written explanation and a list of next steps, instead of a gut-feel opinion.

See Getting Started for the full walkthrough.

Project Moneyball is the company behind Performance Radar, founded by Avner Baruch. Its mission is to help revenue teams “Scale Faster, Better & More Effectively” — turning real calls into objective, coachable signal instead of opinion. Performance Radar is its call-scoring product.

Learn more at projectmoneyball.com.

Sales and customer-success leaders and managers who coach reps, enablement teams measuring whether training moved the needle, and reps who want objective feedback on real calls — not opinions.

Four steps: Import a transcript (click Import transcript or drop a file on the Message Center), review the expected minimums and weights, click Run, then read the score, analysis, and recommendations. You can Download PDF for a clean branded report to share.

The full walkthrough is in Getting Started.

A plain-text transcript: .txt, .vtt, .srt, .md, .log, or .csv. Export the transcript from your call recorder or notetaker and import the file here.

Not yet — it’s on our roadmap. Today you export the transcript from your notetaker and import the file. Direct integrations that auto-pull transcripts from your notetaker are planned. If you tell us which tool you use (info@projectmoneyball.com), we’ll factor it into prioritization.

One number with its meaning underneath. Green means the call met or beat the standard, amber means it fell a little short, red means a serious gap (or a safety cap, when a critical skill badly missed). Evidence Coverage shows how much of the scorecard the call actually gave evidence to score.

More detail in Getting Started → Reading the score.

A template saves a reusable scorecard — the Desired State (min) and Weight per skill — for a call type. Select one and click Open template (or double-click) to apply it; Save template stores the current setup; Reset restores the call-type default. Your saved templates are private to your account and aren’t visible to anyone else.

We only show the rep’s role when the call gives enough evidence to determine it reliably. If it can’t be determined, we leave it out rather than guess — and a note appears in the Message Center explaining why. On calls with more than one rep on your side, the name and role shown are the call owner — the rep who actually drove the conversation and was scored, not whoever opened the call.

We built the product around data minimization. We do not store your full transcripts — each transcript is processed in memory for a single scoring run and is never saved at rest or written to logs. We keep only the derived analysis (scores, confidence, short evidence excerpts, summaries, recommendations) so your results can be re-displayed without reprocessing the call.

Your transcript is processed in memory by our Cloudflare Worker and sent to Anthropic (Claude) to score the call; the results are assembled, validated by an output gate, and returned to you. What we keep lives on Cloudflare:

• Derived analysis in Cloudflare KV, keyed to a one-way SHA-256 hash of the transcript, kept ≤ 30 days then auto-deleted.
• Your scorecard templates in Cloudflare D1, under your account.
• Usage counters against a hashed email.

Data is encrypted in transit (TLS) and at rest. Access is identity-gated via Cloudflare Access (Zero Trust) — every request carries a signed identity token we verify before any processing. Hosting is on Cloudflare’s US / global edge. See the Security Brief.

We use a small set of sub-processors to run the service, each under a data-protection agreement. We don’t sell your data and don’t use it for advertising.

• Cloudflare — hosting, database, key-value storage, and identity / access.
• Anthropic (Claude) — AI scoring of the transcript. Transcript data sent via the API is not used to train their models.
• A transactional email provider — only if/when you use the email-a-report feature.

We give notice before adding or replacing a sub-processor. Details in the Data Handling Overview.

Data	Stored?	Where / how long
Full call transcript	No	Processed in memory only
Derived analysis (scores, confidence, short evidence excerpts, summaries, recommendations)	Yes	Cloudflare KV, keyed to a one-way hash, ≤ 30 days then auto-deleted
Scorecard templates you create	Yes	Cloudflare D1, under your account, until you delete them
Usage counters	Yes	Cloudflare KV, against a hashed email, rolling short windows
Application logs	Yes	No transcript content or model output; limited operational retention

Yes. Because full transcripts aren’t stored and the derived-analysis cache auto-expires within 30 days, most call data is short-lived by design. Templates and account data are deleted on request or account closure, and you can request a fresh run that ignores and overwrites the cache. Email info@projectmoneyball.com.

Yes. A formal DPA is available for execution before any paid use, and we’re happy to complete security questionnaires and share our sub-processors’ compliance reports (Cloudflare, Anthropic — both SOC 2 audited) on request. Email info@projectmoneyball.com.