Skip to main content
Scorecards are the foundation of Craft. Get them right and the AI becomes your best coaching tool. Get them wrong and scores feel random. This guide covers the 5 rules that separate great scorecards from frustrating ones.

​
How Scoring Works

Each scorecard item is scored 0-5 based on what the AI can detect in the conversation:

​
Rule 1: Be Specific and Objective

The more precisely you describe the behavior, the more accurate the scoring.
“Did the rep present multiple options to the customer?”

​
Rule 2: One Behavior Per Step

Keep each step focused on a single behavior. This makes scoring accurate and coaching actionable.
  • Bad: “Present options, explain financing, and handle objections”
  • Good: Three separate steps — Options Presentation, Financing Explanation, Objection Handling
When you combine behaviors, a rep who nails one but misses another gets an averaged score that helps no one.

​
Rule 3: Focus on Business-Driving Behaviors

Ask yourself: “Does this behavior directly impact close rate, average ticket, or customer satisfaction?” High-impact (include these):
  • Options presentation
  • Objection handling
  • Value positioning
  • Financing discussion
  • Close attempt
Low-impact (skip these):
  • Generic small talk
  • Rapport building (unless specific and measurable)
  • Administrative tasks

​
Rule 4: Be Explicit About Timing

The AI doesn’t enforce order based on step sequence. If timing matters, say so in the criteria.
  • Bad: “Check in with customer during the middle of the conversation”
  • Good: “Does the rep talk to the customer DURING the inspection? (after starting the inspection but before finishing it)“

​
Rule 5: Use Observable Criteria

The AI can only score what it can detect in the transcript. Avoid subjective language. Test: Ask yourself “Could two people reading the same transcript agree on whether this happened?” If not, make it more specific.

​
Advanced: Weights and N/A

​
Scorecard Weights

Each item has a weight (default 1) that controls how much it affects the overall score. Bump high-impact steps to weight 2-3 so they move the needle more. See Scorecard Weights for the full math.

​
N/A Conditions

Allow the AI to mark a step as “Not Applicable” when it doesn’t apply. Always specify what triggers N/A. Example: “Explain membership benefits → Mark N/A if customer states they are already a member or rep references their existing membership.”
Create separate steps with N/A conditions rather than one step with branching logic. It’s cleaner and scores more accurately.

​
Example: In-Home Sales Scorecard

Start with 5-8 steps. You can always add more later. Fewer steps = more accurate scoring and clearer coaching conversations.