Introducing Qualified Signal's AI Predictive Model

Qualified Signals is a new account-based buyer intent product. Visit the Qualified blog to see how ‍Signals surfaces the purchase intent of users exploring your website.

Tooba Durraze
Tooba Durraze
February 1, 2022
min read
Apple Podcast LinkGoogle Podcast LinkSpotify Podcast Link
Apple Podcast LinkGoogle Podcast LinkSpotify Podcast Link

What is the Qualified Signals AI predictive model?

At Qualified, we’re firm believers in trust and transparency. Our customers deserve total clarity into how our product is designed and how their data is handled. When we announced Qualified Signals, our new account-based buyer intent product, we knew transparency would be crucial. This is especially true given the nature of the product.

Signals is an AI product that unlocks first-party website engagement data and uses a proprietary scoring algorithm to predict the purchase intent of your target accounts. Historically, terms like “AI,””scoring,” and “algorithm” have been a bit murky. So we decided it’s time to pull back the curtain and give our customers more confidence in this groundbreaking new product. 

Signals surfaces the purchase intent of accounts exploring your website

The Signals model gives customers visibility into the accounts that are actively exploring their website, utilizing the power of first-party (1P) data. These might be companies that are already on your target account list or others that aren’t even on your radar. The Signals model does this by taking into account various inputs when visitors arrive on a website, actualized as two views:

  1. An aggregated list of high level account information
  2. A detailed view of individual accounts that includes events with backlinks allowing you to see which important events are happening and across which visitor. 

The premise of Signals anchors on a score and a trend, which provide several value propositions to our customers. First, we help customers identify accounts they have not been reaching out to, highlighted as those with high scores but may be excluded from a target account list. Second, we provide a possible prioritization framework for sales reps to understand how much website activity is happening with respect to their target accounts, in which they can determine who to reach out to based on the Signals score. Finally, we provide customers with rich account insights, aggregating data all in one place to surface the important information they want to see. 

The Signals model analyzes hundreds of thousands of website inputs 

The following inputs go into the predictive scoring model. 

  • Mouse moves, clicks and scroll depth: Visitor website activity, including scroll depth down a webpage, signals deeper engagement
  • Page views with high intent: A visitor explores pages across your site, each page holding a varying weight as defined by you, the customer. 
  • Active time on site: A visitor actively views website content, versus idle time, triggering a key intent signal
  • Chatbot conversations: A visitor engages with a chatbot, displaying a heightened interest in learning more.
  • Human conversations:A visitor chats with a sales rep over live chat, showing increased engagement
  • Voice calls: A visitor talks with a sales rep over the phone clearly signaling a high purchase intent
  • Meeting booked:A visitor books a meeting with a sales rep, showing a desire to continue the sales process
  • Visitor recency and frequency: Visitors from one account return repeatedly and often, demonstrating ongoing interest
  • Multiple visitors from an account:Multiple visitors from the same account browse the website, signaling internal discussions

Additionally, each of those dimensions is assigned a weight, determined by how much or how little they contribute towards propensity to buy. The weights normalize and adapt over time allowing for the model to learn and become smarter. These models are also customized to the customer’s organization, by way of being able to assign both high- and low-intent pages. High-intent pages have a positive impact on the score, while low-intent pages depreciate the score. This adaptation allows for the model to be specific to the organization, and not a generic out-of-the-box solution. 

Qualified Signals categorizes each account using purchase intent data

Signals Account Trends show a high-level overview of purchase intent for target accounts

In Signals, accounts are categorized by “purchase intent.” There are one of four possibilities defined by a range of scores, also known as temperatures:

  • Cold (temperature between 0°-19°): accounts that have extremely low engagement with your website.
  • Warm (temperature between 20°-50°): accounts that are beginning to engage with your website. 
  • Hot (temperature between 51°-80°): accounts that are actively engaging with your website and perhaps having some conversations with your reps.
  • On Fire (temperature between 80°-100°): accounts with significant active time on site, a number of different visitors from the same account, are having conversations with your sales reps using Qualified, and should be prioritized for active engagement by your sales team.

These purchase intents correlate with the account temperature score, ranging from 0-100°. Since temperatures can fluctuate in real time, so too can intent. An account can go from cold to warm very quickly. To measure this, we use Account Trends.

  • Cooling (decrease of greater than 10 degrees)
  • Neutral (within 10 degrees)
  • Heating (greater than 10 degree difference)
  • Surging (greater than 25 degree difference)

This shows how Qualified Signals uses an account’s purchase intent data to increase or decrease their score based on their website engagement over the past 14 days. If they are neutral or cooling, it means their website activity has stalled. If an account is surging, it means there’s been a significant spike in recent activity. 

An account that is surging but does not yet have an open Opportunity should be a prime candidate for outbounding activities. A cold account that is suddenly heating or surging should also receive attention. An account that is On Fire, but cooling warrants a different level of attention to help reengage the account and bring them back to a heating or surging trend.

Our scoring model uses sessions as points-of-interest. We think of the model like an object in a contained environment. An object can have flashes of heat applied to it, which causes its temperature to rise. Over time, the object cools to room temperature when nothing is happening. The observed temperature represents a value that we modify to produce the signal score. With this model, we can mimic hotness in the sense that sessions are flashes of heat, and the intensity of the heat depends on the importance of events as deemed within our proprietary scoring model. Additionally, certain sessions should not affect the score, low confidence sessions, so we filter them out. This way we have the most accurate data to predict propensity to buy 

Signals presents purchase intent data to accommodate your sales team's needs

Signals intent also doesn't stop at giving you just single points of aggregated data. The information is presented in a multitude of ways.

SIGNALS ACCOUNT TRENDS VIEW: this is a high-level overview of intent for all accounts over time, categorizing each by trend—like cooling, neutral, heating, or surging—to denote whether purchase intent is climbing or waning. This overview can also be customized using unique Salesforce Account data to hone in on the accounts that matter most, like ABM tier, account owner, region, or industry. 

SIGNALS ACCOUNT 360: At the individual account-level, Signals showcases hundreds of data points about website visitors rolled up under one account. This includes a dynamic graph that visualizes purchase intent fluctuations over time as well as an account timeline—a detailed, highly visual, timestamped overview of notable website events that occurred throughout the buying process. This offers sales reps a robust account profile and gives total visibility of each interaction on the account. 

SIGNALS FOR SALESFORCE: Signals data can also be found inside Salesforce where sales reps spend a significant portion of their day. This data is discoverable on the Salesforce Account record for quick access and pertinent Signals data points can be pushed to custom fields. Revenue Operations teams can then build custom reports and dashboards to guide sales reps and their prospecting strategies.

Overall, Qualified Signals allows you to have timely and actionable insights that only get smarter over time. The combination of Signals and Qualified allows you to leverage the full breadth of possibilities that the scoring model can provide. 

Want to learn more? Check out our top 10 strategies to see how you can put Signals data into action across your sales and marketing organization. If you’re interested in exploring how Signals can drive more pipeline for your business, feel free to chat with us right on the website at any time.

About Tooba

As the Director of Business Intelligence and Strategy at Qualified, Tooba is currently finishing her Algorithmic Data Science PhD from MIT. She specializes in looking at AI models and how they interact with both big and small data to gain insights.

Related articles

Qualified in Action

Quick demo?

Discover how we can help you convert more prospects into pipeline–right from your website.

Contact Us