Deal scoring: why pipeline stage is not a probability, and what to score instead
Stage probability tells you where a deal sits in your process. A deal score reads the evidence on the deal and shows which ones need work this week.
Deal scoring is the practice of giving every open deal a number that reflects the evidence on it: how well the account fits, who on the buyer’s side is involved, how recently they replied or met, whether a next step is booked, and how the deal’s age compares with your normal cycle. A deal score measures the condition of a deal. Pipeline stage only records where it sits in your process.
Most pipelines already carry a percentage on every deal, and most teams read it as a probability. This guide covers why that number misleads, what to score instead, the plays for each score band, and how to check that the score works.
What is a deal score?
A deal score is a single number on an open deal, built from factors that each describe something observable. It is also called an opportunity score or a deal health score. There are two ways to produce one:
- Rule-based. You choose the factors and the weights, and the score is the weighted sum. Anyone can read why a deal scored what it did.
- Predictive. A vendor’s model studies your closed deals and returns a likelihood to win, with a list of the factors that moved it.
Stage probability vs. deal score vs. forecast category
These three sit next to each other on a deal record and answer different questions.
| Stage probability | Deal score | Forecast category | |
|---|---|---|---|
| Question it answers | How far along is this deal in our process? | How strong is the evidence on this deal? | What is the rep committing to? |
| Who sets it | An admin, once, per stage | A model reading deal data | The rep or manager, by judgment |
| What changes it | Moving the deal to another stage | Any change in the underlying data | A person changing it |
| Best used for | Weighted pipeline math across many deals | Ranking deals for review | The forecast call |
| Main weakness | Ignores age, activity, and who is involved | Only as good as the data reps log | Optimism, and sandbagging |
HubSpot’s default deal properties show the difference: deal probability is “updated automatically by HubSpot when a user moves a deal to a new stage,” and forecast category is a pick from Not forecasted, Pipeline, Best case, Commit, or Closed won. Neither reads what is happening inside the deal.
Why stage probability misleads
A stage percentage is an average that someone typed into pipeline settings. It can be a fair estimate for a group of deals and still say very little about any one of them.
| What you see | What it means | What to do |
|---|---|---|
| Two deals in “Proposal,” both at the same percentage | One arrived yesterday and one has sat for 90 days. The percentage cannot tell them apart | Add time in current stage to the review view, and score it |
| Weighted pipeline looks healthy, the quarter still misses | Weighted amount is amount times stage probability, so stalled deals keep counting at full stage value | Compare weighted pipeline with pipeline in your top score bands |
| Close date moved for the third time | Stage and percentage are unchanged | Count pushes and subtract for them |
What to score on a deal
Start with five to seven of these. Property names in the last column are HubSpot’s; other CRMs have equivalents.
| Factor | Why it matters | Where the data lives |
|---|---|---|
| Account fit | A deal with a poor-fit company is harder to win and harder to keep | The fit score on the associated company |
| Buying group coverage | Forrester found that 95% of B2B purchases involve three or more people. One contact on a deal is a risk at any stage | Number of associated contacts, and each contact’s buying role |
| Recency of two-way activity | A reply or a held meeting shows the buyer is still in the deal. Emails sent show only that the rep is | Logged meetings and replies. Last contacted is a rough proxy, since it also counts outbound |
| Next step scheduled | A dated next step with the buyer is the simplest sign a deal is moving | Next activity date, and the next step property |
| Deal age vs. typical cycle, and time in current stage | A deal well past your normal cycle needs a second look, and time in stage catches the deal that advanced quickly and then stopped | Create date against the median cycle on your won deals, and time in current stage |
| Mutual plan and close date pushes | A plan the buyer agreed to is evidence. A close date that keeps moving is evidence the other way | A mutual plan checkbox, and a push counter (usually a custom property set by a workflow) |
| Competitor involvement | A named competitor changes win odds and the work needed | A competitor property on the deal |
| Champion status | Someone inside who sells for you when you are out of the room | A buying role value, or a champion checkbox |
| Security, legal, or procurement engaged | Late-stage evidence. Buyers rarely spend legal time on a deal they plan to drop | Contacts with those roles on the deal, or review date properties |
| Amount vs. typical | A deal several times your usual size tends to take longer and involve more people | Amount, compared with your median won deal |
A missing value should add nothing. If the champion field is blank on most deals, fix the logging before the factor earns a weight.
Negative factors and disqualifiers
Some evidence should pull a score down, whatever else is true. Keep the list short and make each one countable.
| Negative factor | A starting rule | Why |
|---|---|---|
| No two-way activity in N days | No reply or meeting in 21 days, or a third of your typical cycle | Silence is an early sign of a stalled deal |
| Close date pushed repeatedly | Two or more pushes | Each push says the timeline was a guess |
| Single-threaded | One contact on a deal past your first stage | If that person leaves, the deal goes with them |
Set the day counts from your own closed deals.
A worked example
Here is a demo deal in the proposal stage, 62 days old against a typical cycle of about 50 days. Stage probability says 60%, the same as every other proposal-stage deal. The score reads the evidence.
- Account fit 20% 8.4
- Buying group coverage (3 of 5 roles) 20% 6.3
- Two-way activity (reply 4 days ago) 20% 8.7
- Deal age and time in stage 15% 5.6
- Next step scheduled 10% 9.2
- Champion confirmed 10% 7.4
- Close date pushed twice 5% -6.8
The math: 8.4 at 20% is 1.68, 6.3 at 20% is 1.26, 8.7 at 20% is 1.74, 5.6 at 15% is 0.84, 9.2 at 10% is 0.92, and 7.4 at 10% is 0.74, which adds to 7.18. The pushes subtract 6.8 at 5%, or 0.34, for a deal score of 6.8.
The breakdown is more useful than the total. The buyer is replying and a meeting is booked, so the deal is alive. The weak lines are coverage and age: two of five roles have never been in a conversation, the deal is past the usual cycle, and the date has moved twice. The rep’s job this week is to reach the two missing roles.
Plays by score band
Bands turn the number into work. These cutoffs are a starting point; move them once you have validated the score.
| Deal score | What it usually means | Rep play | Manager play |
|---|---|---|---|
| 8.0 and up | Strong evidence throughout | Protect the timeline. Confirm the mutual plan and start paperwork early | Skip in deal review unless the amount is large |
| 6.0 to 7.9 | Alive, with one or two weak factors | Find the lowest factor and fix that one. Most often it is multi-threading | Spend deal review time here, where coaching changes the outcome |
| 4.0 to 5.9 | Evidence is thin or going stale | Get a reply and a dated next step within a week, or move the close date | Ask what has to be true to close this quarter. Pull it from Commit if nobody can say |
| Under 4.0 | Little sign the buyer is still in it | Send a closing-the-file note, then close lost or recycle to nurture | Clear these out before the forecast call so they stop inflating pipeline |
Where to multi-thread: sort deals scoring 6.0 to 7.9 by the coverage factor, lowest first. Those are live deals with one or two people involved.
Using deal scores in pipeline reviews and forecast calls
With a score on every deal, a review can skip the walk through the whole pipeline and go to the deals where the number and the rep’s story disagree.
- Sort by disagreement. Put forecast category and deal score side by side. A Commit deal scoring under 6.0 and a Pipeline deal scoring 8.0 and up are the two conversations worth having.
- Review the breakdown. Ask about the lowest factor. “Who else have we met?” gets further than “How do you feel about this one?”
- Compare two pipeline totals. Show weighted pipeline next to the amount in deals scoring 6.0 and up. A wide gap means stage probability is counting deals the evidence does not support.
- Keep the forecast a human call. The score informs the rep’s category. A rep may know something the CRM does not, and the review is where that gets said and logged.
Deal scoring in HubSpot and Salesforce
Both major CRMs ship a deal score, and they take different approaches.
- HubSpot scoring tool (rule-based). You build the rules. Deals can be scored with Sales Hub, and deal scores are combined fit and engagement scores, per HubSpot’s guide to the lead scoring tool.
- HubSpot deal score (predictive). HubSpot says deal scores “predict the probability of winning the open deals in your pipelines,” are “powered by AI,” and read as a percentage, so 85 means an 85% likelihood. It is listed for Sales Hub Professional and Enterprise, and a deal score card shows key factors and score history.
- Salesforce Einstein Opportunity Scoring (predictive). Salesforce says each opportunity gets a score from 1 to 99, along with the factors that contributed most, positively and negatively.
Predictive scores suit teams with many closed deals and consistent logging who want a likelihood without building a model. A rule-based score suits teams that want to choose the factors, set the weights, and explain every point in a review. Plans change, so check the vendor’s page first.
Common deal scoring mistakes
| Mistake | What it causes | Fix |
|---|---|---|
| Scoring stage as a factor | The score copies stage probability and adds nothing | Leave stage out, or give it a small weight |
| Counting emails sent | Busy reps look like engaged buyers | Count replies and held meetings only |
| No negative factors | Old deals keep high scores forever | Subtract for silence, pushes, and single-threading |
| Calling the score a forecast | Leaders multiply amount by score and trust the total | Use the score to rank and review. Forecast separately |
| Never checking it against outcomes | A confident number that does not separate wins from losses | Validate by band every quarter |
Our piece on where teams get ICP scoring wrong covers the account side.
How to validate a deal score
A deal score earns trust when higher bands win more and slip less.
- Pull closed deals from the last four quarters, won and lost.
- Use the score from a fixed point, such as the day the deal entered your proposal stage. The score on the day a deal closed already knows the answer. If you have not been saving scores, start stamping the score into a property at a set stage.
- Group deals into your score bands and calculate win rate and slip rate for each. Slip rate is the share of deals that closed after their original close date.
- Read the pattern. Win rate should rise with each band and slip rate should fall.
| What you see | What it means | What to do |
|---|---|---|
| Win rate climbs from the lowest band to the highest | The score separates deals | Keep the weights |
| Bands look about the same | The factors are not telling deals apart | Find the factors that differ most between won and lost, and reweight toward them |
| The top band wins, but slips as often as the rest | The score reads strength, not timing | Add weight to pushes, next step, and time in stage |
Where BoostIdeal fits
BoostIdeal is a scoring product where you build the model and set the weights by hand, with AI helping you analyze your data and draft a starting point. It works with HubSpot, or a CSV you upload. Scores are out of 10, and in HubSpot each one writes back to a property with its reasons. BoostIdeal scores records in HubSpot, including leads, contacts, accounts, and custom objects, and can use the fit of the account behind a deal. You can run several models side by side, and let your own HubSpot lists and workflows act on the result. Scores recalculate hourly by default, and in near real time with webhook sync. For the account fit factor, start with our ICP scoring guide and five-step framework.
When it is not the right fit: BoostIdeal does not forecast, and it does not chart how a score has changed, so the point-in-time stamp described above is something you set up in HubSpot. It is also not the right fit if your data is not in HubSpot and you do not want to work from CSV uploads, or if you want a vendor to hand over a number you never question.
Frequently asked questions
What is deal scoring?
Deal scoring gives each open deal a number based on evidence such as account fit, the buyer roles engaged, recent two-way activity, a scheduled next step, and deal age. Sales teams use it to rank deals and focus pipeline reviews.
What is the difference between deal stage probability and a deal score?
Stage probability is set once per pipeline stage, so every deal in that stage shares the same percentage. A deal score is calculated for each deal from its own data, so two deals in the same stage can score very differently.
Is a deal score the same as opportunity scoring?
Yes. Opportunity scoring is the same practice under the name Salesforce uses for its deal object. Deal health score is another common name for it.
Does HubSpot have a deal score?
HubSpot has two. Its scoring tool lets Sales Hub customers build rule-based combined scores for deals, and its predictive deal score uses AI to estimate the likelihood of winning each open deal. Check HubSpot’s knowledge base for current plan requirements.
Should a deal score replace the sales forecast?
No. A deal score ranks deals by the strength of their evidence. A forecast is a commitment about amount and timing that a person makes. Use the score to challenge forecast categories.
How do you validate a deal score?
Take closed deals from the last four quarters, group them by the score they had at a fixed point in the deal, and compare win rate and slip rate by band. Win rate should rise and slip rate should fall as the score goes up.
The takeaway
Pipeline stage describes your process, and its percentage is the same for every deal in that stage. A deal score reads what is true of each deal: who is involved, when the buyer last responded, what is scheduled, and how long it has been open. Score a handful of factors you trust, subtract for silence and pushes, review the deals where the score and the forecast category disagree, and check every quarter that higher scores win more often.