Pitchbase lets sales reps practice on AI-simulated calls. This page aggregates, anonymously, what the first 600 analyzed simulations show: where reps fail most, and how fast practice makes them improve. This is our data, not a third-party study, and we own its limits (see below).
Where B2B reps fail the most
According to Pitchbase internal data, the average starting score on a cold call is around 35/100: cold calling remains the hardest exercise. Broken down by skill (on sessions scored axis by axis), two areas stand out as the weakest: closing and objection handling.
| Skill (Sales DNA) | Average score /100 |
|---|---|
| Qualification | 39 |
| Opening | 37 |
| Objection handling | 33 |
| Closing | 31 |
In other words: reps open and qualify decently, but stall when handling an objection and closing. These two axes are where targeted practice has the most impact.
Practice drives fast improvement
The clearest signal in our data: repetition pays off, fast. For users who run at least 3 scenarios, the average score improves by +40% in about one week. These users run an average of 5 sessions of ~3.6 minutes each.
Key figures
Methodology and limits
We publish these figures with their limits, because transparency beats false precision.
- Samples: 600 simulations total; 74 sessions scored axis by axis (granular scores); 39 regular users for the improvement measure. Small samples: these are early signals, not a generalizable truth.
- Score definition: 0 to 100, scored automatically by AI (GPT-4.1-mini, multi-pass transcript analysis). Details in the methodology.
- Improvement: gap between a user's first and last sessions, over ~8 days. Known bias: self-selection (motivated users practice more), no control group.
- Scope: 95% of sessions are cold calls, so the figures mostly apply to cold prospecting.
- Privacy: aggregated, anonymized data, no individual identifier. Security & GDPR.
Data refreshed quarterly. Last updated: June 2026.
Practice on your weak spots
Pitchbase runs realistic AI buyers to train objection handling and closing, with scored feedback after every call.
Book a demo