Sales, RevOps, and finance teams use sales forecasting to predict how much revenue their company will generate over an upcoming period. Forecasts are built based on deals already in the pipeline, past sales performance, or both, and help to set sales quotas, plan headcount and spend, and catch problems early.

There are tools that can help you run a sales forecast, but the forecasting method you use is more important than the tool. A powerful platform still produces a bad number if it's running the wrong method for your data.

There are different methods to run a sales forecast. A team with two years of clean historical data and a stable sales cycle needs a different method than one launching a new product with no sales history. And one method alone isn’t normally enough, since every method has a blind spot. Running two or three side by side means one method's weakness gets caught by another.

This guide covers 12 sales forecasting methods with formulas and worked examples. There’s a comparison table to match methods to your situation, and a framework for choosing (and combining) the right ones.

What is Sales Forecasting?

Sales forecasting is the process of estimating future sales revenue for a set period, based on pipeline data, historical performance, or a mix of both.

Sales teams use forecasts to set sales quotas, plan headcount, and build territory plans. Finance teams use them to estimate what commission expense will look like before the actual payouts are calculated. A faulty forecast throws off every one of these decisions, and by the time the error shows up, quotas are already set, headcount is already hired, and expense projections are already wrong.

Sales Forecasting Methods Compared

The table below gives a side-by-side view of what data each method needs and who it fits best.

The 12 Sales Forecasting Methods

The following 12 methods fall into four groups: quantitative, pipeline-based, judgmental, and hybrid. Each one fits a different combination of data availability and sales motion. 

Quantitative

Quantitative methods forecast purely from numbers like past sales, deal values, and growth rates. They work best when a team has enough clean historical data to trust the pattern, and they struggle when the market shifts faster than the historical trend can account for.

Historical Trend

Historical trend forecasting projects future sales by applying past growth rates to future periods.

Say a company closed $400,000 in Q1, $440,000 in Q2, $480,000 in Q3, and $520,000 in Q4, a steady $40,000 increase each quarter. Extending that trend forward puts next Q1 at $560,000.

The limitation: This method assumes the future looks like the past. It breaks down the moment something changes, such as a new competitor entering the market, a pricing shift, or a product launch. The historical data cannot account for these events.

Best for: Teams with at least a year of stable, consistent sales history and no major changes expected in the near term.

Moving Average

Moving average forecasting takes sales from a set number of recent periods, averages them, and uses that average as the forecast for the next period.

Moving average = Sum of sales for the last N periods ÷ N

‍Example: Say a company wants a three-month moving average, and the last three months brought in $300,000, $340,000, and $320,000. Adding those and dividing by three gives a forecast of $320,000 for the next month. The NIST/SEMATECH Handbook of Statistical Methods notes that a moving average treats every period in the window equally, unlike methods that weight recent data more heavily.

The limitation: A moving average reacts slowly. A sudden jump or drop in sales gets diluted by the older periods still sitting in the average, so the forecast lags behind what's happening.

Best for: Teams with steady, low-volatility sales who want a simple method that smooths out normal month-to-month noise.

Exponential Smoothing

Exponential smoothing forecasts future sales by weighting recent periods more heavily than older ones, so a shift in the last month or quarter moves the forecast more than a shift from a year ago.

Forecast = α × (Actual sales, most recent period) + (1 − α) × (Previous forecast), where α is a smoothing constant between 0 and 1

Example: A company's exponential smoothing forecast for last month was $500,000, actual sales came in at $550,000, and the team uses a smoothing constant of 0.3. The new forecast is (0.3 × $550,000) + (0.7 × $500,000), or $515,000. A higher smoothing constant makes the forecast react faster to recent changes; a lower one keeps it closer to the historical pattern.

The limitation: Picking the smoothing constant takes some trial and error. Teams usually test a few values against past sales and keep whichever gives the most accurate result, or start around 0.2 to 0.3 and adjust from there. Get it wrong, and the forecast reacts too much or too little to real changes.

Best for: Teams with several periods of sales history who want a forecast that adjusts faster than a moving average when recent performance changes.

Regression-Based Forecasting

Regression-based forecasting estimates future sales as a function of one or more variables that influence revenue, like marketing spend, headcount, or lead volume.

Single-driver regression uses: Sales = a + b × (Driver variable) 

Multi-driver regression extends the same idea to several variables at once: Sales = a + b₁X₁ + b₂X₂ + ... 

Example: Historical data shows sales average $3,000 with no ad spend, and every $1,000 spent on ads adds roughly $2,000 in sales. The formula becomes Sales = 3,000 + (2 × Ad spend). A $10,000 ad campaign would forecast to $23,000 in sales. A multi-driver version might add headcount or lead volume as additional terms in the same equation.

The limitation: Regression assumes the relationship between the driver and sales stays stable, and it needs a large enough historical dataset to calculate a reliable relationship in the first place. Add too many variables without enough data points, and the formula starts treating random coincidences in the historical numbers as if they were real relationships, which makes it look accurate on past data but produces unreliable forecasts going forward.

‍Best for: Teams with clean historical data on the specific variables driving sales, and enough data points to calculate a reliable relationship.

Pipeline-Based

Pipeline-based methods obtain a forecast from deals currently active in the CRM. They analyze things like the stage they’re in and how quickly they move through the pipeline. While they accurately reflect the current reality, they're only as reliable as the pipeline data feeding them.

Opportunity Stage

Opportunity stage forecasting assigns each open deal a probability of closing based on its current pipeline stage, then sums the weighted values into a forecast.

Forecast = Σ (Deal value × Stage probability)

Example: A team has three open deals: a $50,000 deal at the proposal stage (60% probability), a $30,000 deal at the negotiation stage (80% probability), and a $20,000 deal at the discovery stage (20% probability). The forecast is (50,000 × 0.6) + (30,000 × 0.8) + (20,000 × 0.2), which comes to $58,000.

The limitation: The probabilities are usually set once per stage and applied to every deal. So, an unusually strong or weak deal gets the same weighting as an average one at that stage. The forecast is only as accurate as those probabilities, and as to how consistently reps move deals through stages.

Best for: Teams with a defined sales process and clean CRM data, where stage progression reliably reflects how close a deal actually is to closing.

Pipeline Forecasting

Pipeline forecasting estimates future revenue from the total value and composition of open opportunities. It uses pipeline coverage as the key measure of whether there's enough pipeline to hit the target.

‍Pipeline coverage ratio = Total pipeline value ÷ Sales target

‍Example: Take a team with a $900,000 sales target for the quarter and $2,700,000 in open pipeline. The coverage ratio is 3x, in line with the three-to-six times coverage that's typically considered healthy. A team sitting at 1.5x coverage against the same target has a real gap to close, regardless of what the deals in that thinner pipeline are individually worth.

The limitation: Coverage ratio tells a team whether there's enough pipeline in theory, not whether that pipeline will actually convert. A team can hit healthy coverage with a pipeline full of poorly qualified deals and still miss the forecast.

‍Best for: Teams that want an early, directional read on whether they have enough pipeline in play, used alongside a more granular method like Opportunity Stage.

Length of Sales Cycle

Length of sales cycle forecasting estimates when open deals will close by applying the average time deals take to move from open to closed-won.

Example: If a team's deals average 45 days from open to close, and a deal opened 20 days ago, that deal is forecast to close in roughly 25 more days. Applied across the whole pipeline, this method estimates a close date for every open deal, so a team can project revenue by week or month, not just by quarter.

The limitation: An average sales cycle hides a lot of variation. A $200,000 enterprise deal and a $5,000 self-serve deal rarely move at the same speed, so applying one average cycle length across a mixed pipeline can put close dates in the wrong quarter.

Best for: Teams with a fairly uniform deal size and sales motion, who need to forecast when revenue lands, not just how much.

Lead-Driven

Lead-driven forecasting projects revenue from the number of leads entering the funnel, multiplied by the rate at which leads historically convert to closed deals and the average deal size.

Forecast = Number of leads × Lead-to-close rate × Average deal size

Example: This quarter, a team generates 500 leads, converts 8% of leads to closed deals, and closes an average deal size of $4,000. The forecast is 500 × 0.08 × $4,000, or $160,000.

The limitation: This method assumes lead quality and conversion rates remain consistent. A spike in lead volume from a new marketing channel can inflate the forecast if those leads convert at a lower rate than the historical average used in the formula.

Best for: Teams with high lead volume and a well-established funnel, where enough historical leads have moved through to calculate a reliable conversion rate.

Judgmental

Teams rely on human expertise in judgmental methods, so they’re the go-to method when there isn't enough data to run a quantitative or pipeline-based method. The problem with judgmental methods is human bias. Optimism, pressure to look good, or plain gut feeling can skew the numbers, and there's no formula to catch it.

Delphi Method

The Delphi method builds a forecast by collecting anonymous rounds of expert estimates and narrowing them toward consensus over several rounds.

Example: Typically, five to 20 people with relevant expertise, such as sales leaders, product specialists, and market analysts, submit an estimate and a written justification for it. A facilitator shares an anonymous summary of all responses with the group, and each expert revises their estimate in light of what others argued. After two or three rounds, the estimates usually converge. Anonymity is important because it keeps a senior voice in the room from anchoring everyone else's number before the group has reasoned through it independently.

The limitation: It's slow. Running several rounds with a panel of experts takes days or weeks, which rules it out when you need a fast turnaround. It also requires having genuine domain experts available to participate.

Best for: High-stakes, infrequent forecasts, like launching in a new market or pricing a new product, where getting it right matters more than getting it fast.

Sales Force Composite

Sales force composite builds a forecast from the bottom up. Every rep forecasts their own book of business, and managers roll those individual numbers up into a team, region, or company total.

Example: A team of eight reps, each forecasting their own pipeline, might produce individual numbers ranging from $80,000 to $220,000. Add them together, and the roll-up becomes the company forecast for that period.

The limitation: Reps have an incentive to shade their numbers. Some sandbag, forecasting low so a strong quarter looks like they beat expectations. Others can be optimistic, especially if they’re under pressure to appear on track. Either pattern, uncorrected, skews the roll-up in a predictable direction.

Best for: Teams that want rep-level accountability built into the forecast, and that track each rep's forecasting accuracy over time so sandbagging or overoptimism gets caught and corrected.

Intuitive Forecasting

Intuitive forecasting is when a sales leader or executive sets the forecast based on experience and market feel, without a formal model behind the number.

Example: A VP of Sales who has run the same territory for six years might look at the current pipeline, factor in a competitor's recent product launch and a slow month for a key account, and land on a number that no formula produced but that reflects context a spreadsheet wouldn't capture.

The limitation: Intuition is hard to audit and easy to get wrong without noticing, since there's no formula to check it against. Forecasting research generally finds that judgment holds up better when it runs through a structured, repeatable process rather than an unstructured gut call.

Best for: Experienced leaders forecasting in a fast-moving or unusual situation where no historical data or model applies, ideally paired with a second method as a check.

Hybrid

Hybrid methods combine two or more of the approaches above, usually a quantitative or pipeline-based baseline layered with something that adjusts for signals the baseline can't see on its own. 

AI-Powered Predictive Forecasting

AI-powered predictive forecasting uses machine learning to score deals and generate forecasts from a wider range of signals than a person could track manually. These signals include deal stage, engagement patterns, rep behavior, historical outcomes, and more. It is updated continuously as new data comes in.

Example: A model trained on thousands of past deals might learn that opportunities with declining email response rates in the last two weeks close at half the rate of otherwise-similar deals still at the same stage. A rep or sales manager reviewing the same pipeline by eye would likely miss that pattern, since it only shows up across a large dataset. 

The limitation: Predictive sales forecasting like this works best layered on top of clean pipeline data. A small or inconsistent dataset produces an unreliable model. It also takes time and volume to become useful; an early-stage sales team won't have enough historical data for the model to find real patterns yet.

Best for: Teams with a reasonably high deal volume and consistent CRM data, who want to layer pattern detection on top of an existing quantitative or pipeline-based method rather than starting from scratch.

How to Choose the Right Sales Forecasting Method

Twelve methods are a lot to pick from. The following three questions should help you trim your shortlist and choose the right methods for your situation. And remember: No single method above is complete on its own. Combining different ones tends to yield more accurate results.

What Data Do You Have?

If you’ve got less than a year of sales data, you can rule out most quantitative methods since they all need a track record long enough to show a real pattern. Pipeline-based methods work better as they forecast from anything currently found in your CRM. Judgmental methods are also an option if you’ve got the right expertise available to you. Once a few years of consistent history build up, it's worth adding a quantitative method back into the mix, since it removes the judgment and bias the other two groups carry.

How Stable is Your Sales Motion?

If you're selling the same product to the same buyer profile quarter after quarter, trend-based methods hold up well since the pattern reflects a business that isn't changing much. If you're in the middle of a pricing change, a new product launch, or a shift in target market, consider pipeline and judgmental methods instead. These methods react to what's happening in the moment.

Who is the Forecast For?

A single method is fine if your forecast will be used to plan next week's calls, since the stakes are low and you can adjust course quickly if it's off. Use multiple methods if the stakes are higher, like if the forecast is going to the board, or if finance will use it to plan headcount and commission expense. Get it wrong, and the company overhires, understaffs, or misjudges payout costs. You can start with a quantitative baseline, then check it against an Opportunity Stage read on the actual pipeline, so the number reflects the real deals in play. This way, you have evidence to back up the forecast if someone on the board or in finance questions how you got there.

One rule of thumb applies to all questions: don't rely on a single method. Run a second one alongside it, and see if there are any discrepancies, as these will uncover any risks. The right sales forecasting software can run multiple methods side by side automatically and flag an issue the moment it comes up.

How AI Changes Each Forecasting Method

The rise of AI sales forecasting has added a layer of automation and pattern detection to these methods. This means less manual number-crunching and a forecast that updates as new data comes in without waiting for someone to rerun it. The table below shows how AI changes each method depending on the group it belongs to.

FAQ

What are the four types of sales forecasting?

The four types of sales forecasting methods are quantitative, pipeline-based, judgmental, and hybrid. Quantitative methods forecast from historical numbers alone, pipeline-based methods forecast from open deals in the CRM, judgmental methods rely on human expertise when data is thin, and hybrid methods combine two or more methods.

What is the most accurate sales forecasting method?

There isn’t one. Instead, it comes down to data quality and fit. A sophisticated model run on messy CRM data will underperform a simple method run on clean data. The most reliable results generally come from combining a quantitative baseline with pipeline inspection rather than betting everything on one method.

What is the sales forecast formula?

The most common baseline formula is Forecast = Σ (Deal value × Stage probability). You add each open deal's value weighted by its likelihood of closing at its current pipeline stage. This works well with a defined sales process and clean CRM data, but it breaks down when stage probabilities are outdated or applied inconsistently across the team.

How do you forecast sales for a new product?

With no sales history to draw on, quantitative methods don't apply yet. Lead-driven forecasting works if there's already a lead funnel generating interest, and judgmental methods like the Delphi method fill the gap when even lead volume is too thin to forecast from.

What are the steps in sales forecasting?

Five steps cover most of it: pick the methods that fit your data, set the forecast period, gather the inputs each method needs, produce the forecast and reconcile it against a second method, then review it monthly and adjust as new data comes in.

From Forecast to Plan

Forecasts are constantly shifting. They can move mid-quarter when a big deal slips, a new competitor shows up, or the market shifts. When this happens, quotas, capacity, and commission expense projections built on the old number are now outdated.

Most teams don't have a way to move the operating plan when the forecast changes. The forecast lives in one tool, quotas and comp plans live in another, and reconciling them means someone manually updating spreadsheets after every forecast review. CaptivateIQ Planning connects the forecast directly to quota, capacity, and compensation, so when the forecast moves, those numbers move with it instead of drifting out of sync until the next planning cycle.

Book a demo to see how it works.

‍