5 Crucial Considerations When Choosing a B2B Predictive Scoring Vendor
A little background — as a Director of Analytics Solutions at Leadspace, I’ve spent the past 2 years working very closely with B2B marketers on scoping, implementing and measuring predictive scoring models. By spending the time understanding their true needs, I feel confident enough to come up with this list of 5 considerations that should be top-of-mind when embarking on this kind of journey.
Predictive scoring is a highly effective way to focus and align your sales and marketing efforts. Using machine learning, an effective predictive model lets you cut through all the noise to reach only the best possible leads and/or accounts.
That said, in this still relatively young and developing industry, you should be choosing your solution with your eyes wide open.
Here are 5 crucial things to look out for, to ensure you’re selecting the best solution possible for your business:
Robust set of 3rd party signals
When it comes to the firmographic and demographic aspects of a predictive score, your 1st party data is usually limited and/or inaccurate. Search for a vendor that can supplement your 1st party data with a blend of multiple 3rd party data sources and plenty of relevant signals, especially social ones which tend to be the most accurate and up-to date.
Remember that even the best data scientist in the world will not be able to create a good model if the underlying data is unusable.
Integration
The vendor of choice should be able to support real-time scoring to align with your marketing and sales SLAs. Also consider the depth of integration, especially on the 1st party behavioral side.
The ideal vendor should push data quickly into your systems, but also pull all the relevant data to create an optimal scoring model.
Model transparency
Black box models do NOT deliver in the long run. Your sales team will start questioning the scores and if you, as a marketer, won’t have a way to explain why a certain lead or account was scored in a certain way, the whole method will collapse.
Choose a vendor that provides clarity into the different signals and nuances in the model, so you can understand it better and be able to present it internally. Even when the machine learning algorithm is complicated, the outcome should be easy to explain.
Model “interference”
This relates to my previous point regarding the non-black box approach. ‘Interference’ might have a negative connotation, but it’s extremely important (and positive!) in this case. Do NOT let a data science algorithm control 100% of your decision making.
Consider this example: your company used to sell quite successfully to SMBs but made a shift to focus only on Mid-Market and Enterprise. A model that only looks at historic data will probably score SMBs highly — which does not resonate with your business goals. It’s key to choose a vendor that can allow you to introduce rules on top of the data science, to make sure your business objectives are met.
Model refresh
Predictive scoring is a continuous process that should take into account changes over time. This applies to both your underlying data itself — which will inevitably require regular refreshing — as well as your model. Your vendor should refresh the model at least twice a year, based on new deals that are won, deals that are lost, and potentially strategy and targeting shifts.
Don’t compromise on the analytics aspect; there are probably trends in the data that you are not aware of, so make sure the vendor of choice is methodical in providing you with the proper insights.
That’s it, good luck!
Comments? Questions? Let me know in the comments section, on LinkedIn or on Twitter