Queue Analysis & Recommender For Psychic Hotline

Psychic Source has a psychic entertainment and phone chat service that allows members of the service to chat with psychic, or experts, over the phone.

Members may enter a queue when their expert of choice is not available and receive a callback when experts are ready to connect

While in queue, members may enter other queues, speak with other experts, log off, or take other actions.

While member in-queue actions vary, influencing behavior with a good recommendation to shorten or circumemvent the queue may have several positive outcomes for members:

• Increased completed conferences
• Increased total chat time

Well-established methods for constructing recommenders (or 'recommendation engines') may be used to build and test a queue intervention and measure changes in the above business measures.

Major Findings

• There's a positive relationship between queue entries and both the number of conferences completed and total chat time by members; these relationships may be used as a baseline for measuring an intervention

• Queue usage has a positive impact on the business, however 36% of queued calls result in no conference

• Using the current queue flow as a control, we can experiment to see whether queue flows with a recommendation have a more significant positive correlation with number of completed conferences and total minutes chatting with experts

• Usually based on member conference ratings, however it was found that only about 10% of conferences have conference ratings (come back to this later)

• Instead, a behavioral rating metric was built using the frequency a member connects with an expert and duration of calls with that expert (taken from 7% of members with 20 or greater conferences)

• Pre-queue: tests whether members that receive a recommendation instead of waiting in queue perform better on the service
• Beyond the queue: receiving personalized recommendations based on conference history, ratings, or other behavior on website, in carousel, etc

• Implement a pre-queue recommender to see whether number of completed conferences and chat time with experts increases for people that receive a recommendation

• Improve conference rating conversion rate (currently 10%) so that the recommender may be used for more members, newer and existing, in and beyond the queue

• Remove a test member (from 2005) who is still on the service entering a significant number queues