Case Study: 13.3% Conversion Rate
Case Study: B2B Subscription Sales
(5 min read)
Intro:
RallySeller had 62 conversations with new leads through Google Ads, Facebook posts and organic traffic then attempted to warm them up and send them to a landing page to provide payment details for a subscription plan offer.
Results:
| 30.6% of new leads were sent to a landing page as warm leads |
|
13.3% provided payment details and signed up for the subscription offer
|
Analysis:
Here you can see the results from the 62 conversations. The y-axis shows us the score of the dialog:
How Scoring Works
| Score | Details |
| -1 | indicates the bot was ignored |
| 0 | 1st message success and the 2nd message was ignored |
| 1-9 | multiple messages sent and possible call-to-action tap |
| 10 | long dialog and call-to-action tap |
| >10 | indicates multiple call-to-action taps |
19 out of 62 interactions resulted in at least one call-to-action tap (coupled with a “warm up” dialog with the sales agent) after some period of consideration as they acquired info about the offer.
Longer dialogs plus enabling customers to fulfill their need of easy access to information while keeping the conversation relevant with Q-learning resulted in an above average conversion rate (13.3%).
Conclusions:
Although there is an ongoing surge of interest in chatbots there is a lack of data about why people use chatbots research funded by the Human-Chatbot Interaction Design project of Norway found that the most reported motivation for using chatbots was “productivity” specifically in obtaining assistance and information in an easy and convenient manner.
It’s no surprise then that RallySeller’s technique of displaying popular prompts using Q-learning in order to inform and persuade can result in high clickthrough rates (plus longer dialogs) and more call-to-action taps.
The chart above shows major fluctuations however, the overall results are clear: optimally displaying popular prompts encourages clickthrough and warms up subscribers. Any such campaign is likely to be stochastic in that people are unpredictable and we cannot assign with 100% accuracy a probability that someone will respond a certain way just because of another person’s actions. However, with over 20,000 dialog trees being AB tested and probabilities of responses considered each message, RallySeller demonstrates superiority to a simple or basic AB tested flow.
Contact support to discuss your specific use case and unlock the power of reinforcement learning in your next chatbot funnel.