top of page
Search

# AI for Sales - Calculating the Sales Forecast

I mentioned in my first blog on AI for Sales, when you hear the phrase “AI for sales forecasting” it feels like the solution is an algorithm to predict “the sales number”. This is only partially correct. While the ultimate goal is an accurate sales forecast, AI for sales forecasting requires artificial intelligence to be infused throughout the entire sales process. The last 6 blogs have walked through different AI use cases for each of the different areas supporting the sales process to build a foundation for applying AI to create a more accurate sales forecast.

With that background we are ready to review AI use cases for calculating different versions of the sales forecast.

Different Versions of the Sales Forecast

Before we dig into AI for sales forecasting use cases, we need to level set on forecasting in general.

Forecasting is the process to predict or estimate a future event or trend. The process uses multiple methods for estimating outcomes of future events to feel comfortable with a value. As an example, the diagram below shows alternate methods for figuring in on a price to sell your home.

The difference between the method “AI says \$425,000” is the other methods are m

anually performed. The two outer methods, Buyer Valuation and Seller Valuation could be more biased. The inner three valuations focus on different specific methods of forecasting

g. The AI forecast may be more complete because it uses more factors for the estimate and could potentially include factors from all three methods. The importance of the AI forecast is not as much about the number it is forecasting but more about understanding what drives that number. If you understand what drives the outcome you can focus on those tasks to drive it higher. As I’ve mentioned in previous blogs, it is important AI is applied as a “glass box” and not a “black box”.

Current State of Sales forecasting.

Most companies have a forecast method best described as “everybody does their own thing”, within some general frameworks of sales forecasting. It is a fragile labor-intensive process, managed by one person and completed in a way they feel most comfortable, it requires a number of different technologies like CRM application, data extracts, and Excel.

The general frameworks they create their forecast with fits into one or all of the following frameworks:

• Weighted Pipeline – Apply a closing probability to “live” opportunities by pipeline stage. The sales forecast for a given quarter equals the amount of the opportunities closing in the quarter multiplied by their closing probability by stage.

• Forecast Categories – In this framework the sales rep identifies the sales opportunities they will “commit” to closing this sales period and which other sales opportunities could also close if this quarter hits “best case”.

• Time Series Forecasting – this method forecasts future events based on known past events. It assumes there is a cyclical pattern to sales. Time series generally assumes patterns of spending and number of sales reps added will be consistent with past trends. Time Series Forecasting of sales for Business-to-Business companies has limited applicability because it ignores things like changes in marketing spending, changes in company focus or partnerships, accelerated market growth and adding sales reps. An enhanced method of time series – “autoregressive Integrated Moving Average” allows for the inclusion of other outside independent variables.

These methods are significant because as we apply Artificial Intelligence, we use these forecasting frameworks. The difference is AI will be providing the observations, opportunity selection and percentages based on past sales patterns in a systematic, unbiased manner to create a sales pipeline driven forecast.

Three Use Cases for Numerically Calculating the Sales Forecast

The following three use cases help define additional automated forecast methods for a pipeline-based sales forecast. The table below identifies key challenges creating a sales forecast and maps them to use cases where AI can be applied to help.

Challenge

Use Case

• Understanding the impact of opportunities not yet created at the beginning of the sales cycle that will close by the end of the sales cycle.

• Apply AI to understand opportunities generated and closed in the same sales period based on past sales patterns.

• Understanding sales rep biases vs actual track record for predicting what they are going to close in the period on a sales category/rep commit type forecast.

• Use AI to assist in the generation of the sales rep forecast (category forecast) by providing benchmarks and patterns from past sales periods that suggest: the percent to use to mark up or down the sales rep forecast based on historic sales rep biases; when an opportunity fits the profile of a “rep commit” but hasn’t been categorized by the sales rep that way.

• Understanding how time degrades the weighted sale pipeline forecast.

• Provide an AI based weighted [stage based] based on past sales and time patterns.

The following provide more information on each of the three use cases.

The first use case helps to understanding the impact of opportunities not yet created at the beginning of the sales cycle that will close by the end of the sales cycle. It does this by using the past history of close won and close lost opportunities that were created during the same sales period they closed in. It models the expected sales using “Autoregressive Time Series”. Autoregressive time series recognizes there are:

• External activities such as marketing campaigns, events, number of sales reps, etc. that represent additional variables that can drive sales

• A cyclical pattern to sales made up of a multi period trend, sales cycle in a sales period and random noise in the process.

The results of these calculations can be interpreted and managed via the following two graphs.

The above graph shows the expected sales to be generated in this sales period attributable to leads generated in this sales period and how it decreases over time. The graph on the left is dependent on creating a targeted number of leads in the quarter. The above graph shows the targeted leads to be created (dash line) and the actual leads created (solid line).

The forecast from future sales leads is added to the AI based sales pipeline forecast to arrive at the total forecast for the sales period.

The second use case is to apply AI rigor to assist in the generation of a category-based forecast. It is AI assisted because AI is making two recommendations with regard “Rep Commit” and “Best Case”. As was mentioned above in the “Current State of Sales Forecasting” section, the Forecast Categories method is where you ask the sales rep to identify their sales opportunities they will “commit” to closing this sales period and which other sales opportunities could also close if this quarter hits “best case”.

AI is used to make two recommendations with regard “Rep Commit” and “Best Case”:

• The first recommendation is highlighted in yellow under the “AI Schedule to win Rate” column in the diagram below. This is the AI derived “recommendation” based on past history of the overall accuracy of the rep commit and best-case estimates from the sales rep.

• The second recommendation is the row Commit Suggestions. This is the AI calculation suggesting that some opportunities categorized as Engaged or Stalled should be moved up to the Rep Commit stage.

The table below identifies a sample “forecast category” spreadsheet. The column highlighted in green allows the person doing the forecast to discount the rep commit based on their feel for the opportunities and where the forecast will end up. They fill in this column by referring to the AI Scheduled to win rate column and personal experience.

Summary

This blog covered two use cases for applying AI for calculating the numeric sales forecast. Beyond the use cases of defining the forecast, two key requirements of applying AI for Sales are:

Understanding what drives the outcome (forecast/opportunity status/ account email domain, etc.) is more important than the actual forecast itself”. Understanding what drives the forecast allow you to make changes to increase the forecast.

Artificial intelligence needs to be infused throughout the entire sales process. Good forecast algorithms on bad data just renders a bad forecast. AI needs to be infused into data collection, marketing, pipeline management, coaching and the sales forecast calculations.