How To Use Machine Learning To Listen to Affluents
It's been some time! We've been busy the last year experimenting and then applying machine learning to the world of wealth management/financial advisors. It was tough and took us some time and in the end, it works pretty well, thanks to our great developers. So what better way to resume blogging than sharing with you about our achievements? As a quick recap, our Social Selling Solution, targeting Financial Advisors & Wealth Managers deliver these four value points:
Identify and profile affluent people in a niche market
Listen to these affluent people in social platforms (blogs, Twitter and indirectly through Instagram, LinkedIn, Facebook and Youtube)
Delivers insights and actionable relationship-building opportunities
Measure how effective the Financial Advisors and Wealth Managers are engaging with their target audience
We experimented with machine learning across the board. In this post, I will focus on #2 and #3 i.e automated curation of social listening.In the past five years, we have used a mix of algorithms and people to deliver listening and curation. We were filtering the river of tweets/posts with a HUGE topic and then a trained human curator would look at these tweets and spot opportunities and insights.It worked great, but came with some limitations:
It takes time
The time increases with your list of targets
Sometimes, the value is in the "linked" article or picture, and the topic filtering was missing these opportunities
Curators, even the best ones, have bad days ;-)
So, as machine learning platforms became easier to use and to scale, we decided to give them a try.The objective was clear: automate the listening of affluents and selection of opportunities for Financial Advisors. We already had a curation methodology and a list of categories. We also had tens of thousands of hand-crafted examples from our years of listening manually to affluents. Our R&D team used new technology capabilities (Amazon ML, Sagemaker, Spacy).The results are amazing.The machine is clearly superior to a human in executing this process, even if from time to time, a "false positive" may slip into the list of opportunities and insights.The volume of opportunities detected by the machine is larger than what human curators used to identify ( average 4x).Some opportunities are easier to spot by a machine. Here are a few examples:
Life Events: We observed that many people share family pictures, and that picture analysis is critical to detect life events.
Local Tweets: there are only a few tweets that are geolocalized. Yet, many tweets/posts/stories carry location information. If, for example, I tweet "I am going to CES" or "I'm attending a Warrior's game tomorrow" or "Great view from the Sisters " it's clear that I will be respectively in Vegas, Oakland or Bend. You can't really find people that can do this for any location and any context. Plug in named entity extraction, combined with a Map API, and Bingo!
Authored content i.e article written by your prospects: this often requires analyzing URLs and cross-checking username. To check "me", you would need to connect my Twitter id @dominiq with this blog ecairn.com/blog.
Events: Mentions or links to event platforms (Eventbrite, Meetup...) are important signals to identify an event.
The machine also has a better memory. One of the most frequent complaints from our users was that they were getting too many "opportunities" to engage with the same targets and not enough for targets that are not very active in social media. This is something we fixed by automating the process.Do you have questions about our technology, projects? Ping me on Linkedin or just leave a comment. and if you want to see how it works, we do offer a free trial!