Why your forecast is wrong, and the query that shows by how much.

The assumption underneath.

Almost every CRM ships default probabilities per stage, and almost every company keeps them. The forecast is that percentage times the deal value, summed.

The assumption is that every deal in a stage behaves like every other deal in that stage. It is testable in one query, and it is almost never tested.

What the gap looked like.

Early stages were roughly honest. Late stages were not. Contract Sent configured at 90%, closing at 54%. The gap widened the later the stage, which is the opposite direction to how confidence usually runs.

On a $4.02M weighted pipeline that is $1.34M of difference. It does not surface in a quarterly review because the forecast and the outcome are looked at months apart by different people.

Run it yourself.

One query against your closed deals. If your configured probabilities match your observed rates, you have learned something worth knowing. If they do not, you now know the size of the gap.

What to do instead of guessing.

Weight a deal by what has actually happened on it rather than by the stage somebody dragged it to. Kaypo scores each deal from buyer behavior, shows the inputs, and back tests the weights against your own closed deals so the number is checkable rather than inherited.

Then it sends the reading to each rep in Slack every morning, which is the part that decides whether any of it changes behavior.

Configured probability against observed win rate, per stage

select stage,
       count(*)                                   as closed_deals,
       round(avg(probability)::numeric, 1)        as configured_pct,
       round(100.0 * sum(case when is_won then 1 else 0 end)
             / nullif(count(*), 0), 1)            as actual_pct
  from deals
 where closed_at >= now() - interval '12 months'
   and stage is not null
 group by stage
having count(*) >= 10
 order by configured_pct desc;

Questions

Related reading