But what is the best way to go about creating this model in order to ensure the best results?
This article will outline a few tips and best practices based on some frequently asked questions:
How many events should I have in my scoring model?
Great question. You are probably tracking dozens for events in your product, but generally speaking, there are probably only 10-20 events that really matter to assessing engagement. We have seen successful scoring models with as little as 5-7 events and some with as many as 20+. Seems like 10-20 is a pretty good rule of thumb.
Do be aware that too few events is problematic for creating a good scoring model. If you only use 1, 2, or 3 events, there will likely not be enough diversity in the usage to create a wide enough range to create a meaningful ranking of users/accounts.
How should I weigh each event?
A properly configured engagement score should include a mix of:
Advanced events- Expose and elevate your "power users." This means your "power user" features should be given the highest weight (7 - 10) in your model. These are the features that are not very frequently used, but used by your most advanced and experienced users.
Common events- those triggered by every (or the majority) of your users - such as
Login- should be given the lowest weight (1 - 3) or no weight at all.
Core events- Features that fit in the middle - what you might consider your "core" features - should be given medium weights (4 - 6).
What if I've missed any important events that should be added to my scoring model?
One of the most unique features of Sherlock's Events tab is the ability to identify the events that are driving the most engagement. A few days after creating your first score, you should check this tab - toggle to the
Unscored events and look for engaging events that you might want to add to your scoring profile.
How rigid is this score? Do I only have one shot to make it accurate or can I edit it later without effecting the results?
It is very easy to update and iterate on your Sherlock score. In fact, during your first few days with Sherlock, you will likely be iterating on your score quite a bit. When you change or update your scores in Sherlock, Sherlock quickly refreshes the calculations that drive all your reporting data.
Can I score certain user groups differently?
Yes. With Sherlock, you can create multiple scoring profiles.
We highly recommend maintaining a "general" score for all your users before adding more specific scoring profiles.
Remember, no one knows your product as well as you do. Because of this, creating your scoring model should be very intuitive. In general, you should trust your instincts here. Our team can definitely help with more specific suggestions for you product. If you'd like to schedule a chat, simple send a message using the messenger bubble.