Outreach at the 2019 Spark+AI Summit
If someone told you that the best way to determine a restaurant’s success was by simply measuring the number of people who walked in the door, you’d be skeptical. How many of those who walked in actually stayed? How much did they spend per meal? And maybe most importantly, do they intend to come back?
As consumers, we're not swayed by misleading stats, so why should we as sales reps? At Outreach, we're not. We know that it would be foolish to count all entrances into a restaurant equally, and it would be equally foolish to count all engagements equally. In fact, we’re using Amplify and our investment in machine learning to turn the old adage that reply rates are the most meaningful metric on its head.
From Counting to Classification
Let’s face it: the reply rate doesn’t provide enough detail. Sure, maybe you got a lot of replies, but what if a lot of those replies are unsubscribe requests or negative replies? You would need to know more than just the total number. Instead, what if we went beyond just the reply rate by focusing on intent of the reply: unsubscribe, objection, and positive.
Using Data to Drive Deals
At last week’s Spark+AI Summit conference in San Francisco, Yong Liu, Outreach’s Principal Data Scientist, discussed our work on intent classification. Yong discussed how not all replies are created equal, and how Outreach is using Natural Language Processing (NLP) to train a computer to sort through replies and analyze the customer’s behavior and intent. Categories include:
- Unsubscribe: Please remove me from your email list.
- Objection: Thanks for reaching out, but we’re not interested at this time.
- Positive: I’d be interested in talking on Friday. Do you have time around 10am?
With machine learning, Yong demonstrated how you can accurately measure the performance of email templates and sequences, A/B test at scale, and then clearly and confidently pick out the most successful content based on the reply intent, rather than the reply count. With Amplify's data-driven insights, your team will be equipped to send out the sequences most likely to perform and send the right content at the right time to your prospects.
Doing Better with Less!
Yong and the rest of the Data Science team are continuing their research on intent by developing new machine learning techniques to classify intent with greater accuracy and fewer examples. This work will give you the insights to sell more effectively and close deals quicker!
Next stop: Yong will be in Cypress on May 14th at the High Performance Machine Learning Workshop to present their paper titled, “An Evaluation of Transfer Learning for Classifying Sales Engagement Emails at Large Scale.” Stayed tuned for more updates!