Predictive Intelligent Targeting is an analytical tool that determines a visitor’s likelihood to convert on a website, based on a scoring system which operates according to the visitors predicted behavior. By using a behavioral learning model and scoring system, Predictive Intelligent Targeting ensures that the most relevant customers are targeted and sent an invitation to engage at the optimum time.
  • By default this feature is not provisioned. Please contact LivePerson Support to enable this feature.
  • In order for Predictive Intelligent Targeting to work, the Predictive Dialer feature also needs to be enabled, as these mechanisms  work in conjunction with each other.


Predictive Intelligent Targeting’s self-learning model evaluates the behavior of every visitor on your website and uses algorithms to rate and score each visitor in order to measure its viability to automate engagements.
A visitor’s behavioral profile is built of data like: website activity, visitor profile, geolocation, CRM referrals, social media referrals, and chat history. The system then rates the visitor’s intent and their value; and then scores these visitors on a scale of 0 to 1 (1 being the highest) by the probability that the visitor will convert when a chat assisted invitation is sent.
For example, a low score rating would be attributed to visitors whose behavior and classification show that they have the potential to enter a website to perform a self-service conversion or those visitors who do not have an intention to convert.
Accuracy is maintained by the system by building and comparing multiple models, and through the continual learning and adaptation of these models over time. In addition, in order to have the most revised data on visitor activity on a website, the visitors' rates are recalculated every two seconds.
emptyStringNOTE: It can take up to seven days for the self-learning model to process enough data on the behavior of your website visitors in order to start rating and engaging them.
The Predictive Intelligent Targeting’s model works in conjunction with the Predictive Dialer which is the mechanism that controls the number of invitations sent to website visitors, based on the number of agents and their availability. Predictive Intelligent Targeting scores visitors and assesses which is the most likely to convert when engaged; and Predictive Dialer decides how many visitors can be invited depending on the availability of slots (giving priority to the visitors with the highest score).
The mutual operation of these two capabilities ensures that optimum timing (milliseconds) is used to send out an engagement, as this can be the difference between a positive or negative experience. For example, there may be 1000 visitors on different phases of their journey within a website and only 30 available slots, it is the responsibility of the Predictive Intelligent Targeting to automatically and collectively decide which are the optimum visitors to invite according to the data gathered; and the responsibility of the Predictive Dialer to decide which available agents handle these visitors.

Practical Example

Below is a practical example of how Predictive Intelligent Targeting can make split second decisions based on its real time evaluation capability:
  • Kelly is from San Francisco and has accessed your website from her PC where she has added three items to her shopping cart. After her behavior is rated by the Predictive Intelligent Targeting she receives a score of 0.5 as her behavior implies that she is a self-serve visitor.
  • John is from London and is using his mobile phone to access your website where he has browsed eight pages of your site without performing any action. He is rated with a score of 0.75 as his behavior implies that he may need some assistance with a purchase.
  • Suddenly Kelly has removed two items from her shopping cart and her score increases to 0.8 as her new behavior indicates that there may be an issue which requires immediate assistance.
  • One agent has become available to accept a chat, so Kelly is instantly invited to chat based on the Predictive Dialer’s calculation of agent availability.
    However if the agent was available just before Kelly removed her items, John would be the one to receive the invitation.


This section lists the limitations surrounding Predictive Intelligent Targeting and some suggested ways to work around these constraints.Single channel: Currently Predictive Intelligent Targeting only works with the chat channel. However future development is planned to engage across all channels.
Generated Traffic:  In order for Predictive Intelligent Targeting to work with sufficient analytical data and to generate a substantial learning phase, a website needs to generate at least 200 self-service conversions per week.
Single goal: Predictive Intelligent Targeting is currently able to support a single goal or conversion indicator as the basis for scoring visitors across an account or session. However accounts with multiple lines of business (LOB) can also benefit from Predictive Intelligent Targeting. Below are 2 use cases and approaches for deploying Predictive Intelligent Targeting in these types of accounts:
Use Case 1:
Multiple lines of business with a common goal.
An electronics retailer has multiple lines of business which are categorized by product category across their website. The agents’ skills and visitor traffic is segmented by LOB, where each LOB uses the same conversion variable, but they all equate to the same goal: a product sale or entrance to the shopping cart page. In this way Predictive Intelligent Targeting learns how to convert visitors according to the goal and it is not connected to the defined LOBs.
Score all visitors across the site using Predictive Intelligent Targeting, but manage the experience and agents they are routed to with invitation rules, Predictive Dialer and skills.
Use case 2:
Multiple lines of business, with different goals in the same account, for example: sales and services.
A software company with a website containing both an online store and a service section is managed by two different LOBs. The sales LOB is interested in driving proactive engagements to improve online sales, the service LOB is focused on onsite engagements (buttons) with proactive engagements reserved for other specific use cases. In such a situation Predictive Intelligent Targeting can be defined to be used for one LOB, but not the other.
Score all visitors across the site using Predictive Intelligent Targeting, but manage the experience and agents they are routed to with invitation rules, Predictive Dialer and skills.
Manually created rules:  Manually created rules can be used in conjunction with Predictive Intelligent Targeting, however there may be implications on performance. This is because unlike rules Predictive Intelligent Targeting has a singular defined goal for engaging visitors. Manual rules are limited to selecting visitors based on specific use cases. This may be needed in certain circumstances, for example, during the launch of a new product where there may be strategic reasons to engage these visitors as a priority; or for workarounds when working with a single goal when balancing between multiple LOBs, as described above.
As there may be certain trade-offs which may potentially impact how Predictive Intelligent Targeting performs, we strongly advise that you contact LivePerson Support in order to discuss your requirements.