Guest post by Jingcong Zhao is currently leading Content Marketing at Socedo
No matter how hard you try, no lead scoring model will ever be perfect. Combining data into a series of rules to determine how valuable a lead is to your business seems like a long shot, especially when we start talking about the uncertain realm of inferred, or behavioral, data.
To create a successful lead scoring model, you need to get sales and marketing to agree on what exactly a Marketing Qualified Lead is. Add to this the fact that the buying cycle is becoming increasingly complex, it can be challenging to sort through the available data and create a useful lead scoring model.
And yet, many companies are seeing great success with behavioral lead scoring.
Unfortunately, for every company seeing success, there are others who struggle to make lead scoring work for their business. The ways many marketers get tripped up usually fall into two categories:
- Data collection
- Synthesizing data into a model
Here are two mistakes many marketers make when collecting data for lead scoring and three mistakes when building a lead scoring model:
Mistake #1: Using Only 1st Party Engagement Data
Most marketers today are incorporating behavioral data into their models, and engagements with your owned digital channels, through website visits, emails, content downloads on your websites, and clicks on links to your social media posts, are some of the easiest and most important to collect. Unfortunately, these actions will only capture the leads who are already considering your brand.
With the average buyer conducting 60% of their research before reaching out to a business, the odds are you’ll only be able to collect 1st party engagement data on a small percent of your leads. So, this data should only make up a portion of your lead scoring data.
What is 3rd Party Intent Data?
Unfortunately, people do not go to your website and download your e-books at the very beginning of their research process. Instead, they start the buying journey by doing research on the broader web – by reading industry publications, following industry influencers on social media, and looking at product reviews written by their peers. This type of third party data is known as intent data.
3rd party intent data is any action taken by a lead on the broader web – outside of your owned digital properties – that signals their buying intent and helps inform sales and marketing on how and when to message them.
3rd party intent data can be found in many places, including:
- Social media
- Publisher sites
- Analyst reports
- Company website
- Buying guides sites/software review sites
- Video learning communities (i.e. BrightTalk)
- Print media
One of the most available and useful forms of 3rd party intent data is social media activities. Why?
72% of B2B buyers use social media to research business solutions.
This research – in addition to their normal social media interactions – can provide you with a number of insights into their pain points and stage in the buyer’s journey.
Up until this point, marketers only used social data for general market research and for creating buyer personas, but not for scoring individual leads. The barrier was that social media activities were only available at an aggregate level. Now, with recent advances in technology, social media activities can be matched to real user identities and associated with other contact information.
Social media also offers many more data points to add to your lead scoring model. After measuring our own lead scoring over a period of time, Socedo found that on average, for every 1.4 email clicks and 3.1 website page visits, a lead in our database took 6.5 relevant, industry-related social actions.
Mistake #2: Not Investing in Clean Data
Dirty or unclean data is loosely defined as poorly entered, improperly maintained, non-normalized or out-of-date information.
Dirty data is a rampant problem. Twenty-five percent of the average business-to-business database is inaccurate, according to a study by SiriusDecisions, and 60 percent of businesses reported “unreliable” data health.
There are several causes of this troubling trend, and behavioral data presents its own unique set of problems. The data we collect from our own properties and 3rd parties is often unstructured, such as the text of a comment someone wrote online or the content of the website where a lead was doing research, but we need to capture it in a useable way in our database. Furthermore, outdated behavioral data is just as unusable as outdated demographic data. The fact that someone visited a competitor’s blog a year ago is just as meaningless as an undeliverable email address.
Without clean, accurate data, you cannot develop an effective lead scoring model.
To clean up your database, consider the following tactics.
- Standardize your processes. As your database grows, know the source of your information. You have a lot of control over your own website, emails, and forms, but when it comes to 3rd party vendors ask them how often they refresh their data, whether they will overwrite your existing fields or add their own activities, and in what format the data will enter your system. When in doubt, create separate fields for 3rd party data until you trust it.
- Track what you can use. More isn’t necessarily better. The goal here is to capture repeatable values that can be measured against each other and easily identified, so while a long string of text will be difficult to work with, a keyword or general topic from that text is useful information. Similarly, a list of all six webpages a lead visited on your website isn’t helpful; the fact that they visited three pages per day for two consecutive days is gold. In general, integer and single-word picklist or string fields will always win over longer text fields or even multiple fields.
Mistake #3: Not Having Separate Scores for Behavioral and Demographic Values
Now you’ve got demographic data and behavioral data (both 1st party and 3rd party intent data) in your marketing automation database, you’ll just need to mash them together, right? Not exactly.
You need to keep separate scores for fit and behavior. You’ll need both types of lead score values to be able to distinguish the CEO with little to no interest in your solution versus the lower-level individual contributor with a high interest. Both these leads are valuable but require different approaches.
In the graphic below you can see this lead scoring model represented by two separate gates. The first gate, the fit gate, is based on demographic criteria. Those that are a good fit should then be nurtured and sent your marketing messages. Based on their interaction with that nurturing and your tracking of the lead activities we mentioned earlier, you can decide how they will move through the second gate, based on their behavioral actions.
Mistake #4: Ignoring Score Inflation
One issue many lead scoring models have is score inflation. When you award points to a prospect for some specific behavior, the score needs to be reset after a time period. Otherwise, a lead who is casually engaging with your brand over a long period of time could generate a high lead score that indicates they are ready to buy when in fact they are months or even a year from making a decision.
The value of someone downloading an e-book last week, for example, will be much less significant a year from now. However, many lead scoring schemas don’t reflect the degradation of score value over time, and the result is score inflation. A way to fix that is to use a system of expiration dates and subtract points over time.
You should also be cautious about over-scoring repetitive actions. Since actions like social media activity occur much more frequently, they can run up a lead score.
Mistake #5: Not Iterating on Your Lead Scoring Model
When you develop a lead scoring model using a simple approach, you’re intuiting on what actions matter most and are most indicative of interest. You’re going to get some of it wrong.
You started by scoring certain types of content (whitepapers and case studies) with higher value than other types (eBooks or product sheets). But should they be scored that way?
The only way to know for sure is to look at your conversion data.
Every month or so, you’ll want to run an analysis to see how each attribute or action is correlated to a conversion and the revenue amount associated to leads who took that action. Then adjust the points you give to leads based on your analysis, collect more information, and repeat.
At Socedo we run an analysis of every action included in our lead scoring model and look at the number of times the action was tracked, the lead to opportunity rate, the opportunity win rate, close rate and revenue generated. For example, what percentage of leads who engaged with your brand on Twitter in the last 60 or 90 days became an Opportunity? What was the revenue generated from leads who did this? What about leads who engaged with one of your competitors on Twitter in the last 60 or 90 days? Is the conversion rate higher or lower than your baseline conversion rate?
This allows us to determine in real dollar values how each action should be weighted.
How to Calculate a Basic Lead Score:
- Calculate the lead-to-customer conversion rate of all your leads
- Pick and choose attributes that you believe are common amongst higher quality leads
- Calculate the individual close rates of each of those attributes
- Compare the close rates of each attribute with your baseline close rate, and assign point values proportionally
If you want to get more advanced and make your lead scoring model more accurate, you should use a data mining technique, such as logistic linear regression which automatically calculates the probability that a lead will close into a customer.
Lead scoring is meant to help you prioritize your time with your best, most interested leads. While new trends in ABM and predictive analytics have placed a higher emphasis on a named account strategy, and more and more companies are moving towards an outbound sales and marketing strategy, your sales reps still need to communicate with individual people at the end of the day. Make their time worthwhile with behavioral lead scoring that incorporates intent data.
By keeping your data clean, relevant, and actionable, you’ll already be on your way to lead scoring success.
Have you run into any other problems developing your lead scoring model? Feel free to reach out!
Jingcong Zhao is currently leading Content Marketing at Socedo, a software company that helps B2B marketers reach their target audiences by leveraging intent data from the social web. Learn more about Socedo at www.socedo.com