How to Use Click Stream Analysis to Optimize your Company’s Social Outreach
In this blog, I’ll be discussing how I expanded the recommendation demo provided in Talend’s Big Data Sandbox to influence my promotional Twitter campaign.
Enterprises are now taking data-oriented approaches when defining their social strategy as they find new and interesting influencers around their business. It is critical to implement plans that utilize this information to optimize social cadence, enabling companies to stay top of mind without blowing their marketing budget. A great example of this is the work that was done over at Molson Coors Brewing Co. By correlating their brand outreach to specific weather conditions, they were able to increase the visibility of their posts by 93% while reducing the cost-per-click by 67% compared to their generic ads.
With these analytical approaches taking hold within some our customers, I wanted to use Talend to accomplish something similar.
As mentioned, I did a bit of hacking on the recommendation demo that is part of the Talend Sandbox to complete the job below. For those of you who haven’t tried out our big data sandbox, you can find it here. In the meantime, here’s a little background on the use case. In the example, Talend is collecting clickstream information in real time from a cycling e-commerce store. This information is then routed through a recommendation engine that scores the likelihood of the visitor purchasing an item. Results are then displayed on a web page for analysis.
My addition identifies the product grouping (frames, components, misc) that is driving the most interest on the site and highlights them on social media. Specifically, it creates a tweet that includes a promotional code that could be used to convert visitors from viewers to purchasers. A screen shot of my mapping is below.
The job collects the scored views from the Cassandra table that the previous steps in the demo populate. It filters out lower level scores and then joins the relevant items with their product group. I then average the groups to determine which group had the highest chance of being purchased, sort the information and then find the most relevant group. Finally, I join the main flow with a file containing the preformatted tweets I want to send out.
If I wanted to take the example further, like the Molson Coors use case mentioned above, I could add weather data, locational-based events or additional social feeds to enrich the decision and message of my tweet.
All in all, pretty simple with the most time-consuming part being the Twitter app setup.
Here’s a quick walkthrough on how to do that and set up the tTwitterOutput Component.
Whether you’re sourcing or targeting Twitter, the first thing you need to do is go to https://apps.twitter.com and create an app.
Fill out the application details and go into the application settings to find your keys, you’ll need them to set up the tTwitterOutput component.
Click on manage keys and access tokens which will get you the consumer secret key as seen below.
Drop down to the Access token section and generate your tokens.
Once you have the keys, just fill them into the corresponding fields of the tTwitterOutput Component.
With that, you’re off to the races creating tweets using Talend.
I hope this gives you an idea or two on things you could do with Talend as well as our big data sandbox. If you make any modifications and want to share what you did, we’d love to hear about it. Feel free to tweet @Nick_Piette and @Talend about your story. I’m sure there is some swag somewhere around here that I can send your way. :)