Predictive Dialing Campaigns with Inference React
Our VP Product, Santosh Kulkarni, presented a webinar on the release of our new predictive dialer today. This is currently in open beta. Predictive dialers are the rolls-royce of campaign dialers and clearly the favorites when it comes to dialer technology as they have the potential to significantly increase agent talk time and boost call center productivity. As the name suggests, predictive dialing uses mathematical algorithms to predict when an agent will be free and relies on the principle of dialing ahead. They make use of sophisticated pacing algorithms that can compute how many outbound calls should be placed such that every answered call can be connected to a free agent. There are four critical factors when it comes to the performance of a predictive dialer:
- The pacing algorithm - The pacing algorithm automatically monitors activity in the call center network and calculates the number of calls that should be dialed. Many predictive dialers rely on sophisticated mathematical models that learn as a campaign progresses. If the algorithm is effective, you'll see marked improvement in agent utilisation, and a corresponding improvement in the speed of running campaigns.
- Maximum abandoned rate - One of the critical factors that directly affects the effectiveness of predictive dialers is the abandoned call rate. A predictive dialer needs to be told how aggressive it should be - which is best measured by how often the dialer makes a mistake - connecting a campaign targets too quickly and having them spend too many precious seconds before being connected to a free agent causing them to abandon the call. The goal of a good predictive dialer is to minimise the number of abandoned calls and at the same time keep the agent idle time low. At the very least, dialers should allow call center operators to specify a maximum abandoned call rate for a given campaign in-line with regulatory requirements.
- Campaign and call center size - Given that predictive dialers work on predictions based on statistical data, the more agents you have in a call center, the better the prediction. Dialers that are based on a learning model perform better with large campaigns. This is because they have a bigger data set to learn from and improve their prediction algorithm over the course of the campaign.
- Agent awareness - A predictive dialer needs to know when to get the next call on the line, and when it has left it too late to start connecting calls. Dialers therefore need to be agent aware as they require information such as total active agents, idle agents, and average agent call duration in real time.
Inference's React platform allows BroadSoft service providers to offer enterprise hosted services with the CTI integration required to deliver true predictive dialing capabilities. Inference's Predictive Dialer application on React provides the machine learning intelligence to allow enterprises using BroadSoft to make a step forward in the productivity of their teams running large campaigns.