Sales forecasting software: what to fix before you buy it

Sales forecasting software can make a forecast faster, more visible and less dependent on one person's judgement. It can't make a forecast trustworthy if the data underneath it isn't. Most tools, and every AI forecasting tool, learn from your own CRM history. So check the inputs first: what your stages mean, who owns close dates, how clean the pipeline is and how much usable history you have. Then decide what the tool needs to do.

This guide is for a CRO or sales director who doesn't trust the forecast and is weighing up a forecasting tool. It's fair to software. Tools help, and some teams are ready for them now. The aim is to make sure you buy one for the right problem.

What sales forecasting software does, and what it can't

Forecasting tools do three useful things. They roll up the pipeline automatically, so nobody rebuilds the forecast in a spreadsheet each week. They show which deals have moved, slipped or gone quiet. And the predictive ones estimate the outcome from patterns in your past deals. Microsoft's documentation for Dynamics 365, for example, says its predictions use historical close rates and the current pipeline, and are recalculated every seven days.

What they can't do is decide what your data means. Salesforce Help says plainly that its standard forecasts "rely on how picklist values map between the Stage field and the Forecast Category field". HubSpot's default forecast is a weighted amount: "the Amount multiplied by the Deal probability", with the probability set for each stage. If a stage means different things to different sellers, the tool adds up that inconsistency faster rather than fixing it.

The problem is common. A Gartner survey published in 2020 found only 45% of sales leaders and sellers had high confidence in their forecasting accuracy. Gartner named poor data quality as one of the main causes: only 47% of respondents believed their organisation had high-quality data.

What the software needs from you

The vendors are clear about this in their own documentation. The thresholds differ, but the inputs are the same: stage, close date, amount, and enough consistent history of all three.

Notice what this table says about readiness:

• Salesforce: won't show a prediction without enough history.

• Microsoft: warns that test or dummy data skews the model, and leaves out deals that abandon the defined process.

• Gong (an AI forecasting vendor): its own help pages say outdated opportunities, incorrect close dates and inconsistent deal stages can compromise its projections.

The documentation is clear about these limits, so check them against your own CRM before you watch a demo.

Five input checks before you buy

Run these on your own pipeline first. Each one takes a pipeline export and an honest afternoon.

1. Do stages mean the same thing to everyone? Pick ten deals at the same stage and ask what evidence put each one there. If the answers differ, the stage probabilities are guesses. Our guide to sales process stages shows how to define them by exit criteria.

2. Who owns close dates, and do they move? Count how many open deals have had their close date pushed more than once. Frequent pushes mean the tool will learn your optimism, not your buyers' timelines.

3. What does "commit" mean? If two managers define commit differently, a tool that rolls up commit will mix two standards.

4. How clean is the pipeline? Look for duplicates, deals with no amount, deals untouched for months and test records. Our guide to stalled deals covers how to spot the ones that should leave the pipeline.

5. How much usable history do you have? Check whether you have enough closed won and lost deals, with a consistent history of stage, amount and close date, to meet the thresholds above.

If you answer all five well, you're ready for software. If you can't, a tool will mostly measure how inconsistent the inputs are.

Symptom, likely cause, what confirms it

When buying the software is the right call

Buy it when your inputs pass the checks and the problem is effort, speed or visibility. Typical signs:

• managers spend hours each week building the roll-up by hand;

• leaders can't see deal movement between forecast calls;

• you want an independent prediction to challenge the sellers' calls.

The tool also helps more when leadership owns it. Gartner's 2024 survey of 303 sales leaders found that CSO-led analytics were 2.3 times more likely to achieve higher forecast accuracy than analytics led elsewhere. The same survey found 84% of sales leaders said sales analytics had less influence on performance than leadership expected. Poor data quality was among the top three barriers.

Fix the inputs first when the five checks fail, or when the argument is about what the numbers mean rather than how fast you can see them. Buying then usually means paying for a tool to learn from history you'll have to replace. If the disagreement runs wider than sales, into marketing handoffs or capacity, our guide to choosing between a sales team assessment or commercial diagnostic covers which to commission, and Merlin5's Growth Assessor™ is built for the wider question.

Where a velocity baseline fits

Fixing the inputs is the same work as building a trustworthy baseline. Merlin5's Sales Velocity Analysis™ starts there:

1. a named owner for every dataset;

2. stage criteria, forecast categories and revenue definitions agreed formally;

3. duplicates and inconsistencies removed;

4. one dataset the leadership team uses.

It then measures the four drivers in our sales velocity formula guide: opportunities, deal value, win rate and cycle length. That shows which ones are moving before revenue does.

Merlin5's view, set out in its CRM and process FAQ, is that forecasting improves as a result of how opportunities are managed before they reach the forecast, rather than as a separate exercise. That was the order at Source 1 Solutions. Merlin5 began with diagnostics, then designed a sales operating system with improved qualification standards, forecasting disciplines and pipeline governance. Once that's in place, a forecasting tool has sound history to learn from. If you then need a partner to configure it, our guide to choosing a CRM implementation partner covers how to compare proposals.

Where to start

1. Export the open pipeline and run the five input checks above.

2. Compare each stage's probability with that stage's real win rate over the last four quarters.

3. Decide in writing whether your problem is the inputs or the effort. Buy for the second, fix the first.

Frequently asked questions

Will AI forecasting fix an unreliable forecast?

Only if the history it learns from is reliable. AI forecasting tools learn from your past opportunities. If stages, close dates and amounts were inconsistent, the prediction will be too, or the tool will decline to show one.

Does our CRM already have forecasting built in?

Most major CRMs include a forecast that rolls up the pipeline by stage or category, and some offer predictive forecasting under specific licences. Check what you already have before buying a separate tool.

How much history do we need?

It depends on the tool. Salesforce asks for at least 12 months of opportunity history for Einstein Forecasting. Microsoft's premium forecasting can start with more than 10 closed opportunities, but says accuracy rises with volume. Check the vendor's current documentation against your own data.

Is weighted pipeline a forecast?

It's a starting point. Weighted pipeline multiplies each deal's amount by its stage probability. It's only as accurate as those probabilities, so check them against your actual conversion rates.

Sources

1. Salesforce Help, "Considerations for Setting Up Einstein Forecasting", accessed November 2026. https://help.salesforce.com/s/articleView?id=ai.einstein_sales_forecasting_considerations.htm&type=5

2. Salesforce Help, "Understand How Einstein Creates Your Forecasting Prediction", accessed November 2026. https://help.salesforce.com/s/articleView?id=ai.einstein_sales_forecasting_adjustments.htm&type=5

3. Salesforce Help, "Data Requirements for Sales Cloud Einstein", accessed November 2026. https://help.salesforce.com/s/articleView?id=ai.einstein_sales_data_requirements.htm&type=5

4. Salesforce Help, "Manage Opportunity Stage to Forecast Category Mappings for Pipeline Forecasts", accessed November 2026. https://help.salesforce.com/s/articleView?id=sales.faq_forecasts_category_mapping.htm&type=5

5. Microsoft Learn, "Enable premium forecasting", Dynamics 365 Sales, May 2026. https://learn.microsoft.com/en-us/dynamics365/sales/configure-premium-forecasting

6. Microsoft Learn, "Analyze revenue outcome using predictive forecasting", Dynamics 365 Sales, May 2026. https://learn.microsoft.com/en-us/dynamics365/sales/analyze-revenue-outcome-using-predictive-forecasting

7. Microsoft Learn, "Configure predictive opportunity scoring", Dynamics 365 Sales, August 2025. https://learn.microsoft.com/en-us/dynamics365/sales/configure-predictive-opportunity-scoring

8. HubSpot Knowledge Base, "Set up the forecast tool", updated 9 September 2026. https://knowledge.hubspot.com/forecast/set-up-the-forecast-tool

9. Gong Help, "See forecast projection", accessed November 2026. https://help.gong.io/docs/analyze-see-forecast-projection

10. Gartner, "Gartner Says Less Than 50% of Sales Leaders and Sellers Have High Confidence in Forecasting Accuracy", press release, 12 February 2020. https://www.gartner.com/en/newsroom/press-releases/2020-02-12-gartner-says-less-than-50--of-sales-leaders-and-selle

11. Gartner, "Gartner Survey Finds Sales Analytics Has Less Influence on Sales Performance Than What Leadership Expected", press release, 6 February 2024. https://www.gartner.com/en/newsroom/press-releases/2024-02-06-gartner-survey-finds-sales-analytics-has-less-influence-on-sales-performance-than-what-leadership-expected