Engagement Scoring Decay

[Angus_McCann1]

We are trying to come up with a general engagement scoring model in our Marketo Instance.

Right now we are the beginning phases (creating the model and let it run in the background for a month or two before rolling it out to users) and I'm trying to come up with an appropriate score decay model based on inactivity (the filter would be score has not changed in the past 30 days).

The concern is that if a person is super engaged - gains a bunch of points January through June, but then falls completely off in July, we want to be able to make sure that our decay isn't too light - that is to say that we are not still counting a person in Dec as "engaged" who hasn't engaged in 6 months. 

IDEALLY - We have a person's score decrease by say, 10%, for inactivity so the flow step would be "Change Score (*0.9)" but the only operators are addition, subtraction, and just changing it to a specific value.

I'm honestly stumped and I'm at the point where I'm concerned I just might be over thinking this - does anyone have any tips on how they decided an appropriate score decay value?

[Jay_Jiang]

There are many discussions and blog posts on this. But again, scoring is unique for your business. Think about the lead lifecycle and typical sales cycle length for your business. What actions indicate and how long do leads go inactive for before they typically become disengaged?

Create a date or datetime field field to track "last activity" - this is different to last modified, where last activity catches any action that indicates the the lead is still engaging with your business (you only need to allow leads to run through once a day) - e.g. visited webpage, opened/clicked email, filled out form, SFDC activity is logged/updated. Use this field in your inactivity scoring. 

How much to deduct and for what period of time, is really up to you and your understanding of your business.

[Angus_McCann1]

Sorry I misspoke before - since our marketing users are already tracking success in their programs based on the activities we want to capture, the scoring campaigns are going to be based on that.

Score Increasing programs will look like this:

• SL:
  • Program Status was changed to Success
• FLOW:
  • Change Score +X

*Schedule is daily, people can run through as many times as necessary

And the Decay Campaign is going to be:

• SL: (All)
  • Was sent email 3 times in the last 30 days
  • Last Program Success was in past before 30 days
  • Not Score was changed in the past 29 days
• FLOW:
  • Change Score -[Decay Value]

*Schedule would be daily, but a person can only run through the Decay Campaign 1 time every 30 days.

** I'm aware the SL filters already accommodates for this, but this is just a fail-safe 

I've googled and most of the blogs and article have not been very helpful as to how to determine the correct decay value.

So for example, I've predicted that we can expect over the course of the year, the average person's score get to be as high as around 3,000, without any decay. If someone was inactive for a year, we would want their score to dip to an "inactive" range, about 500. Based on that model, we could have a decay of around 130-140/month.

However, if a person stays engaged for 3 years and then falls off, their score would get as high as 9,000. With a decay value of 130/month, it would take almost 5 years for that person's score to drop to 500.

Has anyone else encountered this problem as well or am I over-thinking this? I just can't help but feel there is an obvious solution here that I am missing...

[Jay_Jiang]

When you have no maximum (or no minimum) lead scores, unless all your prospects take similar routes to become customers, you risk losing meaning in your lead scores. I, for one, have no idea what 3000 even means for your business, let alone someone with a score over 9000!

Consider redoing your lead scoring model so you have minimum 0 and maximum 100. We made 60 = MQL.

This way anyone can pretty well guess what a lead with a score of 80 means, or 20 etc.. which also makes decay easier.

Also, with webhooks, you can use any math operation you want for scoring.

[Anonymous]

This is one area that I think is lacking in Marketo. Other MAPs, like Eloqua, have a 100 point cap.

[Jay_Jiang]

Playing the devils advocate here, but what if, for my specific business processes, I need scores to go over 100?

I wouldn't say that Marketo is "lacking" compared to Eloqua, Marketo gives you the option to build it as you'd like. Plus the fact that "score" type custom fields having their own set of functionality is quite useful.

What you're hoping for is more like a on/off switch in admin where you can toggle the core platform functionality with regard to scoring. Alternatively, there are scoring programs you can import into your instance and simply switch on - something to explore if you weren't aware

[Angus_McCann1]

So doing a maximum of 100 really limits the number of activities we can capture and kind of depreciates the value of higher-value actions - we already have a minimum of 11 activities and plan . (Opening an email ,Clicking an Email, Downloading Free Content, Registering for Webinar, Attending a Webinar, Registering for a Live Event. Attending a Live Events, New to the Database, Registering for a NL, Creating an account, Visits a few pages on our site)

We send an email literally every single day of the week, not to mention this is across two brands, so to cap at 100 would definitely not be sufficient or usable. Further, without having a Marketing Engagement score model is it hard to determine where to place that cap, let alone the appropriate decay amount.

I've asked around and a lot of people advised not to have a cap considering all of the activities we have available to our database. Basically I'm still trying to figure out is how to determine appropriate decay score. Should I look at what we think the most a person could score is and divide by Y to determine a score decay? Should we have a default decay of X and then, if they were inactive, have the decay be 2X? Etc? There's gotta be some math behind this that I can take a first stab at coming up with a relative decay score, but everything I'm seeing online is like "pick a value" which is less than helpful...

Also trying to avoid web hooks / developers from getting involved.