Refine scoring criteria

Export every deal that closed, won or lost, in the last quarter. Include the qualification fields you used at the time: budget, timeline, role of the contact, company size and any score you assigned.

If you have fewer than 20 closed deals, use six months. A small sample cannot give firm answers, but it will show you where to look. Put the export in a Google Sheets tab, or use a report in HubSpot or Pipedrive if the fields are already there.

Write each criterion as a yes or no question, one per column. For example: Was budget confirmed? Did the buyer name a deadline? Was the contact a decision maker? Did the company have more than ten employees?

If a criterion is vague, such as "good fit", rewrite it so two people would answer it the same way. A criterion you cannot answer with yes or no cannot be tested.

For each criterion, split the deals into those where it was true and those where it was not. Count the win rate in each group. This is an invented example to show the method, not real data:

• Budget confirmed: won 12 of 15 deals (80 percent). Budget not confirmed: won 4 of 15 (27 percent).
• Company has more than ten employees: won 9 of 18 (50 percent). Ten or fewer: won 7 of 12 (58 percent).

The first criterion predicts wins. The second makes no useful difference, so it is noise. A large gap means the criterion matters. A small gap means you can stop scoring it.

Some of the best signals are not in your model yet. Read the notes of your five best wins and your five most expensive losses. Look for patterns: a certain role on the call, a trigger event such as a new hire, a particular tool already in use, or a pain that came up early.

Add any pattern that shows up in three or more deals as a candidate criterion. Your call recordings help here, as described in Review call recordings monthly.

Make at most two changes at a time: remove one noisy criterion, add one promising one, or move a weight. If you change five things, you will not know which one made the difference.

Keep the total to six criteria or fewer. More than that, and the team stops using the score. The four-fact test in Qualify hard, because a clean no beats a slow maybe is a good floor to build on.

A criterion only works if the call collects it. When you add one, add the matching question to your framework, as in Structure your discovery framework and Prepare qualifying questions. When you drop one, remove its question too, so the call gets shorter.

Then update the CRM field and tell the team in a short message: what changed, why, and from which date.

Write one line for each change: the date, what you changed, the evidence and the date you will check it. Keep these lines in one place, so you can see later what the model looked like and why.

This also protects you from changing back and forth. If someone wants to restore a criterion, they need to show new evidence.

Run the same comparison next quarter on the deals that came in under the new model. Did the win rate of high-scored deals go up? Did the number of qualified meetings that never convert go down? That second question is the focus of Improve discovery-to-qualified ratio.

• Judging a criterion on three deals.
• Keeping a criterion because it feels right, whatever the numbers say.
• Counting only wins, which hides what the lost deals shared.
• Adding a new criterion without a question in the call.
• Changing the model in the middle of a quarter, so you cannot compare.
• Scoring fields that reps fill in after they know the outcome.

• Every criterion has a win rate with and without it in your sheet.
• Your model has six criteria or fewer, and each earned its place.
• Deals with a high score close at a clearly higher rate than deals with a low score.
• Your team can explain why each criterion exists.
• You have a dated list of changes with reasons.