Check Microsoft's four deal-score limits: 40 won and 40 lost deals, 1,500 scored records per month, 10 models, and history deleted after 30 days.
Key takeaways: read four limits before deal scores are turned on
Before you activate a scoring model, verify the four specific constraints listed on the product page. These limits define the data requirements and operational boundaries for the feature.
- Closed-deal minimum: The page states the minimum number of closed deals required before a model can be created. This includes both won and lost opportunities. You must count these deals within the specific period the page defines. If your current count is below this minimum, the model cannot be generated yet.
- Monthly scored-record cap: The page specifies a monthly limit on the number of records that can be scored. According to Microsoft, you'll get 1,500 scored records per month. This cap applies to the volume of data processed in a given month.
- Model count limit: The page defines the maximum number of models you can keep, both published and unpublished. You must check this number to ensure you have space for a new model if you are replacing an old one.
- Subscription history rule: The page explains what happens to your historical data if the subscription ends. You need to know whether the scores and underlying data remain accessible or are removed when the service is cancelled. For recovery steps, see recover deleted sales records.
Check these four items directly on the page that offers the model. Do not assume these limits apply to other tools or vendors. Each limit is specific to the product and plan stated on that page.
If your closed-deal count is below the stated minimum, wait until you have enough history. Do not turn on the score until the page confirms you meet the requirement.
Score-model limits by publisher in one table
The table below lists the specific constraints documented for each publisher regarding deal scoring models. You can verify these limits directly from the cited sources before activating a model.
| Publisher | What you check before activating a scoring model | Rule copied from that publisher | What this rule does not decide |
|---|---|---|---|
| Microsoft | Minimum closed deals and monthly scored record cap | Your organization must have a minimum of 40 won and 40 lost opportunities that were created and closed during the selected period, and you'll get 1,500 scored records per month according to Microsoft | The specific win rate lift or the accuracy of the score for your specific pipeline |
| Microsoft | Maximum number of models and subscription history retention | You can create up to 10 models, both published and unpublished, for different sets of opportunities, and if your Dynamics 365 Sales subscription expires or your org is deleted, the historical data is deleted after 30 days according to Microsoft | Whether the model continues to function after the subscription ends or if data is retained indefinitely |
| HubSpot | Signal processing for new deals | For new deals, a score will be created after all required signals for that deal have been processed according to HubSpot | The historical minimum count of closed deals required to initialize the scoring system |
| Close | Workflow trigger timing | These Workflows only fire when an Opportunity is updated going forward according to Close | Whether the workflow rescores deals that are already sitting in a status before the update |
Check the exact wording on each page to confirm the current limits for your specific plan.
How many closed deals the page requires
Before you turn on a scoring model, verify that your organization meets the specific history threshold stated by the vendor. The documentation for Configure predictive opportunity scoring | Microsoft Learn defines a hard floor for the data required to build the model. Your organization must have a minimum of 40 won and 40 lost opportunities that were created and closed during the selected period, according to Microsoft. If your organization doesn't have at least 40 won and 40 lost opportunities that were created in the time frame identified in the Train with opportunities from the past field, you can't create a scoring model, according to Microsoft.
The phrase created and closed is critical. It means the opportunities must have both originated and reached a final status within the specific window you choose for training. A deal that was created two years ago but closed last month does not count if the selected period is only the last quarter. Similarly, a deal that was created within the period but is still open does not count toward the minimum. You must look at the intersection of creation date and close date. The page requires records that were created and closed within the chosen timeframe.
This rule applies specifically to the product and plan described on the cited page. It does not establish a universal standard for every sales tool. Do not assume that other platforms require the same 40/40 split or the same time-window logic. Each vendor defines its own data requirements, and some may have different thresholds or different definitions of what constitutes a valid training record. The 40 won and 40 lost figure is a documented limit for this specific configuration, not a benchmark for your entire pipeline or a target for your team's performance.
The monthly cap and the model cap
Before you activate the scoring feature, verify the two volume limits that govern how the system processes your pipeline. These caps define the maximum workload the model handles and the number of distinct configurations you can maintain.
The page for Configure predictive opportunity scoring states that you will get 1,500 scored records per month according to Microsoft. This figure represents the monthly volume of records that receive a score. Check your current monthly scored record count against this number.
The same page also defines the maximum number of models you can keep. You can create up to 10 models, both published and unpublished, for different sets of opportunities according to Microsoft. This cap includes both models that are actively scoring deals and those that are saved but not yet published. It applies to the specific product named in the page title.
The 1,500-record cap controls the monthly processing volume, while the 10-model cap controls the number of distinct scoring configurations you can store.
What a workflow or insight does after the score exists
Once the score exists, the system applies it to new records and may trigger automation. The scope of these actions varies by platform.
In Configure predictive opportunity scoring | Microsoft Learn, the model calculates a score of 80 for Opportunity A and 50 for Opportunity B, according to Microsoft. These figures are documented examples, not a standard target for your pipeline. To retrain the model every 15 days, select Retrain automatically, according to Microsoft.
In Predict likelihood to close with deal scores, a score is created for new deals after all required signals for that deal have been processed, according to HubSpot. The higher the score, the greater probability the deal will be won, according to HubSpot. With the introduction of buying signals, predictive deal scores factor in data related to activities, according to HubSpot.
In Automatically set Opportunity Confidence based on pipeline status | Close Help Center, Workflows only fire when an Opportunity is updated going forward, according to Close. This means a workflow does not rewrite deals already sitting in a status. Close Workflows don’t support branching or if/else logic, so one Workflow for each status is the way to go, according to Close. If two Workflows both match the same Opportunity update, the last one to run wins — and your Confidence could be overwritten unexpectedly, according to Close. For the stage rules behind this, see open, won, and lost stage rules.
In Sales Forecasting | Freshsales | Freshsales Suite : Freshsales, Freddy tracks the pattern of your deals and detects the ones that are likely to convert and, therefore, suitable for sales reps to commit, according to Freshsales. Sales reps can choose deals that have been suggested by Freddy as Likely to Close and commit to them, according to Freshsales.
Check your platform’s documentation to confirm whether your automation applies to existing records or only new updates.
The counts you have today
Use the table below to record the deal counts in your pipeline before you enable scoring. The page for Configure predictive opportunity scoring states your organization must have a minimum of 40 won and 40 lost opportunities that were created and closed during the selected period according to Microsoft. Compare your current figures to check if you meet that requirement.
| Metric | Your Current Count | Publisher Minimum | Status |
|---|---|---|---|
| Example | Won and lost deals counted in the selected period | 40 won and 40 lost opportunities | Ready only when both counts meet the minimum |
Illustrative example: one team checks the four limits
A team has 50 won and 40 lost deals created and closed in the chosen period, so it clears the 40 won and 40 lost minimum. It keeps 3 models, published and unpublished, under the cap of 10, so one more fits. Its scored records stay under 1,500 per month. Only then does it turn the scores on.
Today, count your won and lost deals in the period your scoring page names. If your counts are below the minimum, wait until you have enough closed history. Do not turn on the model until the numbers match the requirement stated on the page.
Related guides: Catalog price before the software deal total: 2 checks · File size limits: 8 checks before evidence uploads · Check the buying company on a deal: 5 record limits.
FAQ: Closed deal minimums before a score model is turned on
How many won and lost deals does the page require before a model can be created?
The documentation for Configure predictive opportunity scoring states that your organization must have a minimum of 40 won and 40 lost opportunities that were created and closed during the selected period, according to Microsoft.
How many scored records does the page include in a month?
You will get 1,500 scored records per month, according to Microsoft. This limit is documented for the specific predictive opportunity scoring feature.
How many models can be kept?
You can create up to 10 models, both published and unpublished, according to Microsoft.
What happens to history if the subscription ends?
Historical data is deleted after 30 days if your Dynamics 365 Sales subscription expires or your org is deleted, according to Microsoft.
Does turning on a workflow rescore deals that are already in a status?
No. These Workflows only fire when an Opportunity is updated going forward, according to Close.