Enhanced B2B Data Can Markedly Improve Prospect Scores
Data availability can limit the quality of B2B prospect scores. We are not talking about sample size, which is important. But about the characteristics of the businesses the prospect score is meant to rank.
A client’s customer was using an “off the shelf” application for prospect scoring. This application used a very limited set of firmagraphics to construct its scores.
The customer knew our client had access to a wider array of business site data. So they asked if we could measurably improve the performance of their prospect scores.
The customer used prospect scores to identify high potential prospects (business sites) in a information technology (IT), B2B marketing list.
The state of B2B business data
There are a number of vendors which provide data on the characteristics (firmagraphics) of business sites. Dun and Bradstreet, InfoGroup, Compass, Orb, V12, to name a few, provide data such as number of employees, industry code and revenue.
Data vendors vary in terms of how many data fields they provide, in the amount of “white space” in these fields, as well as in how many business sites they cover. And, of course, in price!
But all data vendors are constrained by the fact that, unlike consumer demographics, business firmagraphics tend to be more limited in number.
Another set of data vendors provide “enhanced” B2B data. In the IT B2B marketing space, vendors like HG Data and Aberdeen provide data on the presence of certain technologies and technology vendors, counts of technology (e.g. PCs) and estimates of IT spend.
Enhanced B2B data coupled with base firmagraphics yields much more fertile ground for prospect scoring than just base firmagraphics alone.
What are B2B prospect scores?
A prospect score is a number assigned to a list of non-customers to be contacted in an outreach campaign . After sorting this list by the score, the non-customers who are the “best” prospects will rise to the top.
These “best” or “high potential” prospects typically resemble current customers. Converting these into actual customers is the job of the sales team. The prospect score just tells them who they should focus on first.
Data analysts typically construct prospect scores through statistical modeling of a sample of customers and non-customers and their characteristics.
Can enhanced B2B data and methodology improve prospect score lift?
Since our client had access to enhanced B2B data along with base firmagraphics, we structured a test for our client’s customer as follows:
- First, construct a prospect score using only the few firmagraphics the customer was currently using;
- Second, construct a 2nd score that used these same firmagraphics plus enhanced B2B data fields such as IT spend, technology counts and indicators of technology presence;
- Third, construct a 3rd score using all available data and paying more careful attention to data preparation and modeling techniques.
To gauge model performance we used “lift”. In short, lift is the degree to which a score identifies more potential customers compared to using some other score or no score at all. We discuss lift in more detail in a prior article.
Findings – base score
Six (6) firmagraphic fields support the base score: enterprise revenue and employees, site employees and type (HQ, branch, standalone), industry and metropolitan area. The lift chart below shows that the resulting score does identify about 75% more customers (compared to using no score) in the top decile.
But the lift chart does not exhibit a smoothly declining lift from the highest to the lowest decile. Nor is the lift decline very pronounced. This suggests that the “discriminatory” ability of the score (i.e. the ability to tell the difference between a customer and a non-customer site) can be improved.
How do you construct a “lift chart”? Start with a list of companies, some are customers and some are not. Build a model and then score this list. Sort the list by the prospect score, from highest to lowest potential. And break the list into deciles. Find the actual occurrence of customers in each decile and express this occurrence in terms of an index. Index values > 1.0 indicate the score does a better job than using no score.
Findings – enhanced score
Various site-specific data fields for presence of technologies, IT spend and technology counts, as well as the base score’s firmagraphics, support the enhanced score. The resulting lift chart is now smoother and exhibits a more pronounced decline in lift from the highest to the lowest decile.
That is, the discriminatory power of the score has markedly improved. In this case, by about 15% in the top decile.
Findings – complete score
The final score uses a more sophisticated data preparation and modeling methodology. And additional firmagraphics. This adds an additional 5% to lift in the top decile.
The lift chart now suggests that, in the top decile, the score is 2x more likely to identify a business site that is similar to a current customer (compared to no score).
Findings – summary
Another way to look at this is to consider what happens over the first 30% of the file, not just the top decile:
- A prospect score using enhanced B2B data fields in addition to base firmagraphics identifies 12% more potential customers (than using base firmagraphics alone);
- Using a more robust methodology can boost this by another 6 percentage points.
Both data and methodology matter
The take away from this exercise? Both data and methodology matter. But methodology will only take you so far.
The greater the number of enhanced B2B data fields available, the more likely the discriminatory power of the model will improve. The caveat, of course, is that these data fields need to be materially different from each other. If they are all highly correlated with each other, then additional “similar” data fields won’t matter.
Of course, your mileage may vary in terms of the impact on lift. But augmenting your marketing data with enhanced B2B data and using an appropriate methodology to construct your prospect scores is likely to yield a positive ROI.