We built a salary negotiation simulator. Here are 5 things we learned.

When we built Commy, we expected to see what every negotiation coach tells you: people are afraid to state a high number.

That's not what we found.

After analyzing 2,000+ salary negotiation practice sessions — real responses from real professionals, scored by AI on clarity, confidence, and assertiveness — the picture is more specific, and more fixable, than "be more confident."

Here's what we learned.

Finding #1: Most people frame their ask as a question

The single most common pattern in low-scoring responses (under 60/100): the request ends in a question mark.

• "I was thinking maybe around $145k?" — Confidence: 38/100
• "Would something like $155k be possible?" — Assertiveness: 31/100
• "I was hoping we could maybe consider $140k?" — Clarity: 44/100

When you phrase a salary ask as a question, you're asking for permission to want something. It signals to the other person that you're not sure you deserve it.

The fix is simple: state a number, then stop talking.

"Based on my research and the value I've built in this role, I'm targeting $155k."

That sentence has no question mark. It has a specific number. It has a reason. That's all you need.

The same response scored 87/100 on our rubric.

Finding #2: "We" kills your achievement stories

This one surprised us.

In behavioral interview practice ("tell me about a time you led a project"), 71% of responses used "we" as the primary subject when describing outcomes.

"We shipped the feature." "We hit the deadline." "We grew revenue by 40%."

The problem: hiring managers and promotion evaluators can't assess your individual contribution when you blur it into a team outcome. They need to know what *you* specifically did.

The shift is uncomfortable for people who are used to collaborative environments. But it matters.

"I led the architecture decision. I pulled in the right engineers. I presented the results to the exec team."

Same story. Completely different impression.

Average score improvement when switching from "we" to "I" language in achievement sections: +14 points on specificity.

Finding #3: Most people know when to stop — they just don't stop

Clarity scores are surprisingly high in our data. The average clarity score across all drills is 68/100.

But assertiveness scores average 54/100 — 14 points lower.

This gap reveals something specific: most people communicate *clearly* but fail to *advocate*. They explain their position well and then immediately offer an escape route.

"I'm targeting $155k — but obviously I understand if that's not possible, I'm flexible, I know the market is tough..."

The moment you pre-empt your own ask with softeners, you've done the other person's negotiating for them.

State your number. Stop talking. Let silence work.

One of our coaches calls this the "comma clause" problem: every strong statement gets immediately followed by a comma and a qualification. The drill that specifically trains against it is one of the most practiced on our platform.

Finding #4: Confidence is about word choice, not tone

We scored responses on four dimensions, and one of the most surprising findings: confidence in written communication has almost nothing to do with the strength or volume of the words chosen, and everything to do with specific hedging patterns.

The hedging words that most reliably tank confidence scores:

• "I think..." (confidence: -8 pts)
• "I feel like..." (confidence: -7 pts)
• "Maybe..." (confidence: -9 pts)
• "Kind of..." (confidence: -11 pts)
• "Sort of..." (confidence: -10 pts)
• "I'm not sure if this is right, but..." (confidence: -15 pts)

People use these hedges to sound humble and collaborative. In low-stakes conversation, this works. In high-stakes negotiation, it reads as uncertainty — and uncertainty invites pushback.

The fix: before a high-stakes conversation, scan your draft for these words. Delete them. Reread it. You'll be surprised how much stronger it sounds without them.

Finding #5: Practice actually works — measurably

This one we expected, and still it's gratifying to see in the data.

Average score after 1 drill: 58/100 Average score after 3 drills: 74/100 Average score after 7 drills: 82/100

The improvement flattens after 7 sessions for most people — not because they've hit a ceiling, but because the specific scenario has been mastered and they need to practice a different one.

The professionals who improve fastest share two traits: 1. They practice the *same scenario* at least twice before moving on 2. They read the feedback carefully and specifically change the thing that was flagged

The ones who plateau read the score, feel good (or bad) about it, and move on without implementing the specific feedback.

Improvement is about iteration, not repetition.

What this means for your next salary negotiation

If you're preparing for a compensation conversation — whether it's a new offer negotiation, an annual review, or a promotion ask — these four adjustments will move your score:

1. State a specific number, no question mark. Phrase your ask as a statement, not a request.
2. Use "I" for your achievements. Your contributions disappear behind "we."
3. Kill the comma clause. State your position. Stop. Don't soften it yourself.
4. Audit your hedging words. Every "maybe," "I think," and "kind of" signals uncertainty.

You can practice this before the real conversation. The demo at commy.app lets you practice any scenario, get scored on these four dimensions, and see exactly which words to change.

No signup required. Results in 60 seconds.

Commy is an AI communication coach for high-stakes workplace conversations. Our scoring rubric was calibrated against human communication coaches across 200+ sessions. Try it free at commy.app/demo.