Prescriptive forecasting

Prescriptive forecasting is a new type of forecasting that relies on the third age of artificial intelligence, using a neural network combined with deep learning on both structured and unstructured data to formulate a forecast.

How did we arrive at this place, and where do we go from here? Let’s examine the commonly used types of forecasting, their limitations, and what the future holds for forecasting.

Read more below.

Request a Free Demo

On no, we ran into an issue submitting this form. Please ensure each field was filled out correctly and resubmit.

If this problem persists, please reach out to us and we will be more than happy to follow-up from there.

We’ll be in touch soon

In the meantime, explore Collective[i] and find answers to frequently asked questions.

CODIE W&FI HP-S HP-W 2021 Data Breakthrough Awards

It's time for accurate sales forecasting

The most powerful forecasting tool from the world’s largest network of sales professionals.

  • Dramatically improve forecast accuracy
  • AI-driven forecasting
  • Increase accountability across selling teams
  • Eliminate inaccurate CRM data
It's time for accurate sales forecasting

Automated daily sales forecasting

Updated daily based on your data and external network factors that influence your sale.

  • Faster predictions
  • Reduce sales cycles and analyze scenarios in minutes
  • Real-time, buyer-specific selling recommendations
  • Increase deal velocity
Automated daily sales forecasting

Eliminate spreadsheets. Automate your CRM.

Eliminate spreadsheets. Eliminate opinions. Get data-driven results with little to no effort.

  • Automated data capture
  • AI-driven forecasting
  • Network enrichment of contacts
  • Networked intelligence - unlock external factors that matter
Eliminate spreadsheets. Automate your CRM.

Work together, win together

Request an invitation to join IntelligenceTM, the world’s first global network of sales professionals.

On no, we ran into an issue submitting this form. Please ensure each field was filled out correctly and resubmit.

If this problem persists, please reach out to us and we will be more than happy to follow-up from there.

We’ll be in touch soon

In the meantime, explore Collective[i] and find answers to frequently asked questions.

Prescriptive forecasting

A survey by Gartner revealed that only 45% of sales leaders and sellers have high confidence in their organization’s forecasting accuracy. Low confidence in sales forecasts limits sellers’ ability to make sound, data-driven decisions. And unwieldy forecasting tools exacerbate that skepticism.

Yet sales forecasting is one of the most important tools for a business: It impacts where companies invest and grow and influences the valuation of the whole business. “It just takes a lot of effort to get a decent forecast,” says Kunal Mehta, principal of portfolio operations at TCV, “and it’s so critical to running the operations and planning for the future.”

How did we arrive at this place, and where do we go from here? Let’s examine the commonly used types of sales forecasting, their limitations, and what the future holds for forecasting.

What are the types of forecasting?

When it comes to forecasting, there are three methodologies that can be employed: opinion, augmented, or prescriptive.

Opinion-based forecasting

Prior to sophisticated statistical models or neural networks, many forecasting methods relied simply on the intuition of the sellers. These opinion-based methods are subject to many biases: For example, the interviewer may ask sellers leading questions that can influence the way they respond. Opinion-based forecasting methods can garner interesting insights but are not a reliable way to forecast sales or determine the next best steps for a business. Types of opinion-based forecasting include:

  • Jury of executives: This method solicits opinions from a number of individual experts, then has each one revise their opinion based on the information from the other experts. After this process has concluded, the final opinions are combined to create a composite forecast.
  • Sales opinions: The opinion of the sales representatives is used to predict future sales. Their hunches are taken into account and used to create numbers for expected revenue.
  • Expert opinions: Similar to a sales opinion, this forecasting method instead relies upon industry experts to create a forecast based on their observations and feelings.

Opinion-based forecasting relies on qualitative data, human intuition, and feeling instead of quantitative data and numbers.

Augmented Forecasting

The next step in forecasting came in two parts: first using historical data to build predictive models, then augmenting that historical data by combining opinion-based and historical methods. Predicting the future by applying statistical models and other formulas to businesses’ past performance bolsters intuition-based forecasting. Augmented forecasting offers insights into past buying patterns and generates sales forecasts that are more accurate than forecasts based on opinion or historical methods alone. Historical forecasting primarily relies on quantitative data to come up with projections of future sales. The types of historical forecasting include:

  • Time series forecasting: These are statistical methods used when there is a long history of data. The data from the past can be used to predict the current rate of sales and to indicate if the rate is increasing or decreasing. “Time series” refers to looking at data in a specific period over several years to identify seasonal patterns, cyclical patterns, data trends, and growth rates.
  • Simple moving average: The simple moving average looks at sections of historical data — such as the years 2011–2012, 2012–2013, 2013–2014, etc. — and takes the averages, then plots those averages on a graph to compare them. This allows forecasters to see the trends visually and make predictions based on what they observe.
  • Exponential smoothing: This technique can be applied to a variety of historical forecasting methods to account for the fact that recent data is more useful than older data. More recent years are assigned a numerical weight that is higher than older years. To calculate these weights, forecasters use exponential functions to assign exponentially decreasing weights over time.
  • Market analysis: A market analysis looks at the future numbers, characteristics, and trends in the target market. Using historical data, forecasters project the growth of the market, the average purchase per customer, and other types of analysis that are relevant to their industry. A standard market analysis will segment potential customers to create specific projections for each market segmentation.
  • Trend projection: This is a graphical method for which forecasters take annual sales data, graph it, and draw a line through it to note the trends.
  • Opportunity stage forecasting: Opportunity stage forecasting is the process of breaking the sales pipeline into different stages and assigning probabilities of how likely it is to close a deal at each stage.

These are just a few examples of the types of historical forecasting methods. Every method has its pros and cons, so combining quantitative and qualitative methods became a way to create more accurate reports and eventually evolved into a new method: augmented forecasting. One study found that sales forecasting accuracy was increased by using both qualitative forecasting (adjustment of sales data by experts) and quantitative forecasting (time series method using a software program).

Augmented forecasting and AI-lite

Sales forecasting evolved further with the onset of artificial intelligence (AI). Machine learning — computer algorithms that analyze data to detect patterns — enhanced augmented forecasting by looking at historical structured data and analyzing it more quickly than humans could to provide additional information to sellers.

For example, think about Spotify — users give feedback on which songs they enjoy, and Spotify keeps track and generates recommendations of music with similar tempo, genre, or instrumentation. The machine learning algorithm continues to refine the suggestions as users give feedback about the recommendations provided.

In augmented forecasting tools, AI-lite typically relies upon a single data source — usually the CRM of the business. Because the data input is limited, so is the data output. Still, because AI-lite can be integrated with the CRM, it has a huge advantage over forecasting methods that relied on spreadsheets and programs in which data had to be manually entered separately. The ability for AI to integrate directly with a CRM created streamlined and more efficient ways to use data.

Prescriptive Forecasting

Prescriptive forecasting is a new type of forecasting that relies on the third age of artificial intelligence, using a neural network combined with deep learning on both structured and unstructured data to formulate a forecast. Deep learning requires a large amount of data, and prescriptive forecasting combines both internal client data and external, multi-sourced data of buyers and the market to give a broader picture of sales activity. Prescriptive forecasting provides actions and recommendations on the next best step sellers should take based on the deep buying signals the AI has perceived.

What are the limitations of traditional forecasting methods?

The limitations of traditional forecasting include subjective reports and a reliance on historical data and a single source of information.

Subjective reports

The limitations of opinion-based forecasting are fairly clear — because these are not data-driven decisions, human bias sneaks in easily, and decisions made on hunches are not typically accurate or reliable.

Reliance on past results

Augmented forecasting relies on past results to predict future performance. However, turning to the investment community for a moment, a study by the S&P Indices examined if the investment strategy of selecting an active fund on the basis of previous outperformance was a sound idea. The study concluded that historical performance is only randomly associated with a future performance and that choosing a fund based on past performance is a misguided strategy.

Similarly, in the sales industry, striving for an “expected” number of sales based on past performance is ineffective. Sellers may waste a lot of time trying to hunt down every last lead and chase every dead end to meet unsubstantiated quotas. Another aspect of historical data is the temptation to provide conservative estimates so sellers don’t have to work as hard to “perform well.”

Historical data also cannot account for unexpected factors. Take the COVID-19 pandemic, for example. No historical data could have predicted a global event of this magnitude. The pandemic has changed a lot of buying behavior and caused a paradigm shift that companies are still trying to understand and get ahead of.

Single source of data

Marketing in particular has focused on this idea that a single source of data is necessary and good. However, when it comes to sales, a single source of data, such as a CRM, is incomplete. Leaving aside that manual data entry likely leads to gaps in a CRM — common data errors plague 91% of organizations — buyers don’t only interact with one business. Accounting for current trends and buying behaviors that are happening worldwide is something that traditional sales forecasting methods simply can’t do.

How has artificial intelligence impacted forecasting?

To understand how we arrived at prescriptive forecasting, it is important to take a look at how artificial intelligence has evolved over the past and how it has been used to create sales forecasts. Essentially, according to Forbes, there are three distinct ages of AI.

1. Hand-crafted knowledge

Artificial intelligence was first named in 1955, and the emphasis was on mimicking human intelligence with rules-based expert systems. AI was based on rules and logic and followed clear instructions. AI machines weren’t able to learn and didn’t deal with uncertainty very well, since there were no clear-cut rules to follow in unclear situations. This first age of AI was not particularly useful for forecasting except as a glorified calculator — sellers put in the formulas and algorithms, and the AI ran them.

2. Statistical learning

It wasn’t until around 2004 that the second age of artificial intelligence began. Instead of programming AI with strict rules to follow, scientists developed statistical models for certain types of problems and situations. They trained the AI to apply those models to data to drive insights and decisions. This type of AI can learn and adapt to different situations if it has been trained to do so. A downside to this type of AI is that its logical capacity is limited, since it doesn’t rely on precise rules, and its solutions tend to be “close enough” rather than perfectly correct.

Another important development in this era of artificial intelligence is that of the neural network. This is where data is put through different computational layers, each of which processes it in a unique way and then sends it to the next layer. This AI can produce incredibly accurate results. For example, neural networks have been able to recognize faces accurately, write beer reviews, transcribe speech, and even control cars and aerial drones.

In general, neural networks are great at complicated tasks with a lot of gray areas, in direct contrast to the first age of AI. However, these types of AI can’t necessarily assign a reason for the decisions that they make.

3. Deep learning and contextual adaptation

The third age of AI is where we find ourselves in the early stages of today, where neural networks are being improved upon and can use data to teach themselves how and why they have arrived at certain conclusions.

For example, a third-age AI system can look at a picture of a dog and say, “This is almost certainly a dog because it has four legs like most animals, is medium-sized, and has shaggy fur like a dog.” A second-age AI system would look at the same picture and say, “There is an 85% chance this is a dog,” basing that on the statistical likelihood instead of coming to the conclusion through a series of logical conclusions.

Neural networks are a type of machine learning that aims to emulate how a human brain functions. By taking interconnected units that process information by responding to external inputs, neural networks can find connections and draw conclusions from unstructured data. Combined with deep learning, this true artificial intelligence can identify complex patterns in large amounts of data on a level that was previously impossible to conceive of.

Modern forecasting tools such as Collective[i] use deep learning and neural networks to provide cutting-edge forecasting abilities that take into account all the benefits of augmented forecasting (statistical models and qualitative and quantitative data), without relying solely on historical data or a single source of information.

Why is prescriptive forecasting the future of forecasting?

By taking the power of neural networks and applying it to sales forecasting, prescriptive forecasting is able to compensate for the limitations of previous forecasting methods to give forecasts that aren’t just an arbitrary target.

Prescriptive forecasting is able to:

  • Give feedback and information about current buying behavior.
  • Use multiple data sources including opinion-based information, historical data, real-time market data, and internal and external data.
  • Create forecasts that enable sellers to make the right decisions at the right time.

Prescriptive forecasting acts as a yardstick that measures a sales team’s performance on a daily basis, then gives an analysis of what next steps are most beneficial to make. Instead of chasing down buyers who may not be ready to purchase, sellers can use prescriptive forecasting to identify who is most likely to buy and who is not. Without neural networks and deep learning capabilities, this innovation would not be possible.

Steven Messer, co-founder of Collective[i], explains that “…when AI starts to learn something, we use something called gradient descent. This is where there is a lot of data and a lot of trial and error. Every time more and more data sets are added, the AI learns, and then all of a sudden it becomes very good at it’s one single job. It gets to the place where it’s so good that humans can’t decipher the difference.” To achieve gradient descent, the AI must train on a large amount of data, after which it will quickly become extremely accurate and insightful.

At Collective[i], prescriptive forecasting is an integral part of our toolkit. With the industry’s first true neural network intelligence forecasting technology, our tools give you the ability to harness the power of AI and deep learning to make the right moves at the right time. We are dedicated to modernizing every aspect of sales and pride ourselves in providing adaptive, probabilistic predictions around revenue that take into account sellers’ actions and buyers’ responses.

We don’t just solve the issue of forecasting, however. With Collective[i]'s Intelligent WriteBackTM, we solve the problem of incomplete and inaccurate CRM data. Our proprietary technology automatically records each interaction between your selling team and buyers, even if they don’t have a CRM license. Intelligence WritebackTM updates contacts and data as deals progress and prospects change roles.

Our artificial intelligence software is true AI. Using prescriptive forecasting to draw insights based on frequently updated and relevant multisourced data, Collective[i] goes beyond providers who rely on your company’s sales history alone to generate insights.

Connect with us today to learn more about prescriptive forecasting and how Collective[i] can help bring your business into the future of forecasting.

Work together, win together

Request an invitation to join IntelligenceTM, the world’s first global network of sales professionals.

On no, we ran into an issue submitting this form. Please ensure each field was filled out correctly and resubmit.

If this problem persists, please reach out to us and we will be more than happy to follow-up from there.

We’ll be in touch soon

In the meantime, explore Collective[i] and find answers to frequently asked questions.