After working for two predictive marketing vendors, I feel I am in a unique position of understanding a thing or two on how best to pick a predictive marketing vendor. If you’re new to the space, it can be difficult to pinpoint key functionality that will align with your company’s sales and marketing objectives. I’ve compiled quick list of the Dos and Donts of evaluating a predictive marketing vendor.
#1. Do establish your company’s needs and initiatives for predictive marketing
Scope out what you are trying to achieve internally first. Is it improving your sales win rate? Improving lead conversion rates and velocity? I’ve identified 6 primary use cases for predictive marketing: sales productivity, marketing campaign effectiveness, sales forecasting accuracy, demand generation, database segmentation, and cross sell / product penetration.
#2. Do evaluate only the vendors that address your company initiatives
Some predictive vendors excel at helping you identify patterns in your anonymous traffic, others at sourcing raw leads for your database. Pick a vendor that aligns with your company initiatives.
#3. Do perform a live test of the scores
The purpose of a predictive marketing vendor is to predict actual events, not just how well they can predict what you have already done ( closed, meetings booked, conversions ). A vendor can easily built reports to show how well they would have scored leads and opportunities in the past leveraging historical data but it is entirely a different thing to score leads in a live application.
#4. Do pick a vendor that has the expertise and best practices to train and onboard your team for success
As this is a fairly new technology, you’ll want hands on guidance on how to integrate the scores in your workflow, train sales reps, and best practices documentation. Look for a vendor that has published guides, webinars, and success stories
#5. Do pick a vendor that supports integration with CRM and marketing automation platforms
You’ve probably spent a lot of time building out your CRM and Marketing Automation systems. Best to pick a predictive vendor that plays well with the systems you already know and use. No need to reinvent the wheel here. Ideally they should have an easy API integration into your CRM and MAP and not require an extra seat nor take up too much at a storage.
And now for the don’ts
#1. Don’t pick a vendor based on the back test alone
Make sure the vendor can perform in a real world application. A vendor that won’t do a live test of their scores is one that probably won’t perform well in real world application.
#2. Don’t rely on a CSV file to build models
CSV files flatten the data and bring forward information in time. When you give the predictive model information that it would normal learn at a later stage in your lead lifecycle, the model wont be able to accurately predict leads earlier in the lifecycle. The best analogy I can give is it’s like Biff in Back to the Future betting on who is going to win the football game with the sports almanac from the future. In short CSV file will give the predictive model information it shouldn’t have at certain time periods for predicting conversions.
#3. Don’t compare two conversion rates from two different sized sample groups
I call this an apples to oranges comparison. One sample can have a 36% conversion rate and the other can have a 4% conversion rate, but without knowing the size of the sample group or the denominator to be exact, comparing the two conversion rates is truly meaningless. Remember back in algebra when we were taught to find the lowest common denominator? Same thing applies here.
#4. Don’t make a decision on too little data
Often times I’ve seen prospects attempt to make an informed decisions when only 3 leads have converted to opportunity. If your average deal length is 6 months and the trial period is only 30 days, you’ll need to pick an earlier KPI in your sales funnel to evaluate performance and make a decisions. Number of booked meetings might be a good KPI to use.
#5. Don’t compare results from two different time periods
There can be other variables at play such as big tradeshows, higher performing sales reps, seasonality that can impact conversion rates. If you are comparing the performance of the predictive model against the week immediately following Dreamforce, you may not have as many meetings booked. Try to compare a similar time frame. The highest impact tests I’ve seen are when the marketing team splits the SDR team in two, assigns the predictively scored leads to one half of the team and the non predictively scored leads to the other half of the team. After two weeks they ran the same test again but without the predictive scored leads to make it a true blind test. Then with those results the marketing team could prove to management and sales that calling the to predictive leads first had value.
