In this guide 6 sections
The data is wrong, not the maths. Look for stale close dates, stage probabilities nobody has tested against history, deals without an owner, and accounts counted in two territories. Each one has a quick check that you can run on a plain pipeline export before the next forecast call with the sales head.
Stale close dates push deals into the wrong month
Nobody revisits a close date. Three months later the deal still shows this month, with no call logged since spring, and the forecast counts it even though the buyer has gone quiet.
Export every open opportunity, sort by last activity date, and flag anything with a close date inside thirty days and no activity for twenty-one. Owners confirm or move it.
A close date in the past should block saving, and a date moved more than twice should ask for a written reason from the owner.
- Close date earlier than today
- No logged call or email for three weeks
- Close date moved more than twice
- Next step field blank
Stage probabilities are guesses until tested
Many teams inherit a table that says proposal means fifty and negotiation means seventy-five, though nobody remembers who set it and the numbers may suit another business entirely.
Test them against your own history. Take last year’s closed opportunities and count how many that reached each stage were later won, because if proposal stage deals close one in four, your table is lying.
Repeat the count each quarter. Buyers change, and so do price lists, discount habits and product mixes, which is why a table from two years ago drifts away from what closes.
Deals without an owner or a next step
When a salesperson leaves, their opportunities often stay assigned to a disabled user, so they remain in the weighted total while nobody chases them.
Filter for opportunities with an inactive owner, a blank next step or no contact attached, then reassign each in a short meeting. Close the dead ones and record why.
A simple rule helps: no opportunity without an owner, a contact, a next step and a date for that step, with every field mandatory.
The same account counted twice
A buyer with plants in two cities may exist as two accounts in two territories, and both salespeople log an opportunity for the same purchase order. The forecast shows two deals. The bank shows one payment.
Search for duplicate account names, shared phone numbers and the same email domain, then merge the accounts before deciding which territory owns the opportunity. Territory rules should name the owner of every multi-site buyer before the quarter starts, so two salespeople never chase one purchase order.
Check quote numbers too. Two open quotes with identical lines for one buyer point to a single deal, even when different owners created them in different weeks.
- Account names that differ by a suffix
- Two contacts sharing one email domain
- Identical quote lines under two owners
Compare forecast to actuals every month
Keep a table of what was forecast at each month start and what closed, with a column naming one reason per gap: slipped, lost, never real or double-counted. Patterns appear within three months.
Bring the table to the monthly sales review and ask each owner to explain only the largest misses, because the small ones are noise. Skip the rest entirely.
Pipeline health checklist
A single-sheet checklist with four blocks: close dates, stage odds, ownership and duplicates, each with the filter to run on the opportunity export. Each block records the count found and the action taken, and a closing table compares forecast with actual closings and names the reason for each gap.
Open a blank worksheet to printQuestions owners ask
How often should stage probabilities be reviewed?
Every quarter at least. Compare each stage with how many of those opportunities were later won, and change the table wherever history disagrees with it.
Should reps be allowed to override the probability?
Allow it with a reason field. An override without a note hides the very judgement you want to learn from, so make the note mandatory whenever the rep changes the figure.
What is a good forecast review cadence?
A short weekly pass on deals closing within thirty days, plus a deeper monthly comparison of forecast against actuals with the reasons for each miss.
Can software fix a bad forecast?
Software can enforce mandatory fields, flag stale dates and detect duplicate accounts, but the team still has to keep dates honest and update stages on real buyer actions.