Summary

A great scorecard produces the score you'd expect when you listen to the call. The secret is simple: break each criterion into specific yes/no checks. In this guide we'll use one criterion, rapport building, to show exactly how it's done. The same pattern works for every skill on your scorecard.

How Hyperbound Scoring Works

Hyperbound scores every criterion as a clear yes/no check. That's a deliberate design choice, and it's what makes your scores reliable:

• Objective. Each check has one defensible answer the AI can verify in the transcript.

• Consistent. The same call scores the same way every time, across thousands of calls and reps.

• Transparent. Reps can see exactly which checks passed, so every point has a reason behind it.

Partial credit comes from how many checks pass. Pass two of four, and you score a 50 you can point to.

Start With a Common Criterion: Rapport

Some scorecards write rapport as a single question:

"Was the rep effective at building rapport with the buyer?"

The problem: "effective" and "rapport" are never defined, so the AI has nothing concrete to verify. It tends to default to a 0, even after a call where the rep clearly connected with the buyer. The rep sees a zero and loses trust in the score.

Why a Good Call Can Still Score 0

Here's the part that surprises people most. A criterion that is only half-met defaults to 0, not 50. If it isn't fully satisfied, it doesn't get counted.

Now scale that up. Say your scorecard has 10 criteria, all written as broad questions like the rapport one above. A rep gets on a call and does a genuinely decent job, about half of what each criterion is looking for. Every single criterion lands short of "fully met," so every one defaults to 0. The overall score is 0, on a call that was actually pretty good.

Common Misses to Watch For

These are the patterns that quietly send a criterion to 0 when a rep did partially well:

If any criterion on your scorecard has one of these, it won't give credit for partial effort. That's the thing to fix

Break It Into Specific Checks

Rapport isn't one moment. It's a sequence of small moves that build trust. So write it as a ladder of checks, each one a behavior the AI can spot in the transcript:

Section: Rapport Building

1. Did the rep attempt to build rapport?

2. Did they open with light, human conversation (weather, traffic, the buyer's day)?

3. Did they find a common connection (school, city, employer, mutual contact)?

4. Did they surface a personal commonality the buyer acknowledged?

Now a rep who opens warmly but never finds a connection scores a 2 of 4, an honest reflection of the call.

Rule of thumb: if a rep does half a good job on rapport, your scorecard should give them a 50, not a 0.

One tip: keep each check to a single behavior. "Found a connection and the buyer acknowledged it" is really two checks. Splitting them lets the rep earn credit for the part they did well.

Good and Bad Scorecard Example

❌ Too broad (scores most calls 0):

Was the rep effective at building rapport with the buyer?

 ✅ Specific (scores calls fairly):

Same call. The broad version scores it a 0. The specific version scores it a 75 and tells the rep exactly what to work on next time.

Apply the Same Pattern Everywhere

Every criterion on your scorecard works just like rapport. For each skill:

1. Name the skill (rapport, discovery, objection handling, next steps).

2. List the 3 to 4 observable behaviors that show the rep did it well.

3. Let the score follow the number of checks that pass.

That's the whole method. Specific behaviors in, fair and trusted scores out.

Quick Checklist

• Does each criterion name a specific, observable behavior?

• Is each skill broken into 3 to 4 checks?

• Can a half-good performance earn partial credit instead of a 0?

• Is each check a single behavior, not several chained together?

• Would the score match what you'd give the call by ear?