Hey Carrie,
I've recently been working with a number of clients on enhancing their scoring capabilities and building some best-in-class scoring platforms (yes, yes, that's the self-sales pitch but it has to be there 
).
I've been looking for that one "great" document on lead scoring for the better part of 5 years in which time I've built dozens of scoring modules using everything from Marketo's default scoring program to scoring programs with 40 different score fields and hundreds of different scoring points. Sadly, I haven't found that one amazing document and I haven't had the time to write it 
If you're looking at a starting point, Justin Gray generally has some really good high level content for helping people getting into the right mindset for starting out on these sorts of adventures.
When it comes to really rolling your sleeves up and getting it done, there's a few tips I can give you to help you with the process of creating great scoring:
1. Identify the role you want scoring to play before you start!
Many people start their scoring projects from the questions "What should I score" and "how many points should it get", this method generally leads to a misalignment between means and ends. By sitting down with the stakeholder and deciding what the goals of scoring are, it helps you define both the method, the objectives, and most importantly to align the scores you're giving to the business goal you're hoping to achieve.
2. Scoring will NEVER be right the first time round.
I remember when I was first getting started with complex scoring models (around 2009), I spent the better part of 2 months buried elbow deep in excel spreadsheets and even managed to get my hold on a copy of SPSS. I was running cluster analyses and regression models and when my friends asked me if I'd like to go out for a drink I'd answer them in P values it got so bad.
The end result of the process was a phenomenal scoring system, over 100 different points being evaluated and oh so many moving parts. It was a masterpiece. It went live, and started running and from day 1 I started to understand that even in all of my analyses there had been some things that I'd overlooked and certain scores were too high, others were too low, others were unnecessary and other things that hadn't even popped up on the radar were more missing!
Whatever approach you take, you should also have a strategy for optimizing scoring over time, because that's the real trick to hitting your stride. You can start simple and start adding factors as you get more feedback.
3. Avoid double scoring
This is a fun response to write because it's taking me right back to my very early days using Marketo. When adding scoring factors to your model you should make sure to think through whether or not the lead is already getting (or losing) points for the activity you're creating. A classic example I've seen many times in models that clients have built is the double scored webinar registration. In this scenario, they use the traditional scoring trigger for "fills out form is any" and give the lead 10 points for filling out a form. Then they start using webinar programs and say "I want to give them points for registering for webinars", so they add "program status changes to webinar > registered" as another score and give them 5 points for the event. Next thing you know, 50% of hot leads are being sent when they register for a webinar (that's generally when I get a phonecall ). You ask "Did you want them to get 15 points for registering for a webinar?" and the answer invariably comes back, "No I wanted them to get 5 points". Well - you can see where this becomes an issue.
Again, the more factors in your scoring model the more you need to make sure you review new factors to make sure you're not creating overlaps so that you can keep tabs on exactly what's going on.
4. Don't "Hyper-Score" key events!
This is a big one and I was actually having a great conversation with Justin Norris in another thread about this. When scoring drives MQLs people are often inclined to use a "Hyper-Score" for that event to make sure the lead is an MQL. This actually skews your data so far that it leads to all sorts of challenges that you can find listed in the discussion MQL set-up. In short though, don't forget that by giving some things HUGE scores, you make everything else look comparatively less attractive to sales.
If you'd like to discuss it further please feel free to reach out at guy@g8technological.com or via linkedin.
Hope this helps
- Guy
.png)