Saturday, September 26, 2009

Lead Scoring - What's All the Fuss About?

buzz this
I just returned from the TargusInfo Online Lead Quality Summit. Just like last year the buzz was all about lead scoring. TargusInfo claims that they did a survey with their education clients this year asking them whether they intended to score leads and 100% of them said “yes”. So, given the high levels of buzz, who is actually doing scoring and what are the results?

Unfortunately, either no one is talking to keep secret their competitive advantage, or very few companies are actually using scoring. I didn’t find anyone at the conference who could tell me they had done anything more than a little bit of testing with scoring. In fact, I found it completely remarkable that there was a panel called “Lead Scoring: Mend the Leaks in Your Lead Pipeline” on which all 3 panelists were thinking about scoring but had not yet scored a single lead! I am hoping that TargusInfo’s conference dedicated solely to the topic of scoring next year is a little more illuminating.

I do think scoring has huge potential for both lead buyers and lead sellers. One of the bright spots of a good TargusInfo conference for me was Matt McLaughlin from CUNet’s explanation of scoring, verification and validation. It was fairly “101” but the explanation was good and clear so I thought I’d plagiarize his work for the benefit of my readers.

Lead Validation, Verification and Scoring

These services are similar, yet distinct services for lead buyers and sellers that represent a continuum of information depth running from validation (being the simplest form of lead quality signal) to scoring. Here is a basic explanation of each:

Lead Validation: This involves scrubbing individual data points to make sure that they are true i.e. whether the phone number is a valid, in-use phone number, whether the zip is a real zip etc. A lot of lead providers actually use this service to eliminate bad/fake leads.

Lead Verification: This process takes multiple points of information and sees whether they belong to the same person. For example, it will check whether someone called Arthur Smith really lives at 123 Main Street and has a telephone number of (100) 200-3000. Since the information that this analysis is typically based upon is not perfect rather than saying whether a lead is good or bad, verification services will give the lead a rating or mini-score associated with how many of the pieces of information provided could be matched together.

Lead Scoring: People mean a lot of different things when they refer to lead scoring. I’m going to refer to what I think is the most valuable form of lead scoring and therefore the type of lead score that will gain dominance… eventually. That is lead scoring to provide a predictive measure of the outcome of a lead. For example scoring a lead from A+ through E- based on how likely that a lead with the attributes that it has will turn into a converted customer.

How Lead Scoring Works

As Matt McLaughlin pointed out a lead score is the consequence of predictive modeling based on the outcomes of other leads and can be based on 3 types of data:

1. Submitted data

That is basically the data in the lead itself. So with a mortgage lead perhaps you could predict how likely a lead is to convert based on the income, gender, location of the customer or look at factors like whether the lead was sent from a freemail address like hotmail versus a corporate account. You might also want to pull the validation and verification data into the scoring model too. That would certainly lend itself to a more accurate model.

2. Appended data

This is the process of matching a lead with hundreds or thousands of data points about that consumer housed in other databases. For instance if I submitted a lead with my name and address, such databases could tell you that I had recently bought a house in 90404 and have recently spent $X on furniture and had my credit pulled 4 times. A lead scoring company takes into account thousands of data points when they create the model and usually find that less than a hundred data points help to predict how good a lead is. They then use those data points in their algorithm for generating each lead score.

3. Source Information

Source information is the trickiest type of data to decide whether it should be included in a scoring model. Source information describes who generated the lead and under what circumstance. For example, your source information could be that Lowermybills generated the ad from a banner advertisement on a Yahoo! Finance page. Typically however, if you are a buyer of leads, lead sellers won’t share that level of information with you and therefore you’ll only be able to say who sold you the lead.

I have found that the negatives outweigh the benefits of including source information. Knowing who sold the lead and including that in a predictive scoring model usually vastly improves the predictive accuracy of your scoring model. Unfortunately it makes it a very blunt instrument in terms of determining who you should buy leads from and to detect when an individual lead provider’s leads change in quality.

What to measure

When you create a lead score you need to determine what you want to predict. The bias of most users of leads is to want to predict a metric that is most closely associated with profitability or “success”. The problem is that as you go towards indicators of profitability (such as conversion) versus indicators of intent (such as contactability) you have less data points and greater time lags associated with the time it takes to get from lead generation to revenue generation by the lead buyer. Thus a metric associated with profitability may make for a less useful model. The following diagram, based on a slide in Matt McLaughlin’s presentation illustrates this:

The Impact of Scoring

My view is that scoring will have a massive impact on the lead industry over time. It will make it easier to be successful but it will also stretch out pricing. As companies are able to accurately discriminate between lead qualities they will be expected to pay an appropriate price for the lead. For example if I can tell you with 80% confidence that lead will close and I know that your net profit for a conversion is $1,000 eventually that lead’s market price is going to gravitate to close to $800. I personally look forward to that time because it will reduce the amount of time good companies waste working poor leads and will lead to a market where those that win will be the companies that have the best sales process and highest quality sales and customer service teams. That’s a world in which lead management becomes even more important than it is today.


  1. The challenge with Lead Scoring is how much of the score (and what it means) to the sales folks. The primary challenge is that sales folks want the "best leads" and if they believe certain leads are better than others they will try and cherry-pick them. As a result, the "lower quality" leads receive less time and attention which creates a self-fulfilling prophesy as not surprisingly, they convert at a lower rate. The best way to address this is to have set teams work on set leads and compensate them accordingly; meaning pay less for the higher scored leads which are supposed to have a higher probability of closing and pay more for the lower scored ones.

    This has worked in some instances in the mortgage world for LendingTree leads vs. short-form ones, but that is because lenders can pay out different commissions accordingly. Probably true for most other businesses but there are a fair amount of restrictions placed on the enrollment counselors at schools, particularly the for-profit ones, so this presents an additional layer of complexity that will need to be addressed. Curious to see how this plays out.

  2. Ed,

    That's a really interesting insight. LOs/enrollment counselors should be paid more commission for working lower quality leads too. I hadn't thought about it but it makes a lot of sense. That way it still makes a lot of sense to buy cheaper, lower-converting, low score leads since your team will work them harder.

    I think that makes better sense than just hiding the score from your sales team members.

  3. hi Nick,
    There are numerous online and campus-based schools who are quality scoring their leads in real time. Beyond the back-end contact center benefits, I think the industry is gradually moving towards a lead right-pricing model, and more intelligent match-making of students with the best fitting schools -- i.e. those where they are most likely to enroll. At the end of the day, it will make for a more efficient CPL advertising marketplace in EDU, and grow the industry overall.

  4. Hi Nick,

    Thanks for coming out to our event and your solid post-show write-up. I do encourage you to attend our Lead Scoring Summit in May for a deeper dive into lead scoring. We plan to dedicate at least 50% (if not more) of the session time to companies that can talk about their experiences utilizing lead scoring/analytics on not only their web leads, but their inbound phone leads as well.

    We've seen some pretty exciting results from companies in not only the EDU space, but automotive and lending spaces as well. So we hope the event in May will be of interest to a large cross-section of online marketers.

    I know you had some questions regarding who is actually doing lead scoring as well. A very valid question. Couple good links below. One is an article by Liberty University and they talk about how they have been using Lead Scoring. The second link is about Sylvan Learning and their use of lead scoring.

    Feel free to email me with any questions and thanks again for joining us.


  5. great 101 explanation of scoring - something that I agree is very very new to most lead buyers..

  6. This is a great article on lead scoring; however, I noticed most of the content is geared toward business to consumer (B2C) marketing, not business to business (B2B) marketing. In B2B marketing, marketers use marketing automation software to automatically score, or qualify, their leads for the sales people. With rich APIs available marketers also append data, similar to the mortgage example above; however, it's usually information on the business. If you're a marketer with a B2B solution your website is already producing leads for you, you've just got to unlock those leads by using automated lead scoring. There were some comments about identifying "bad/fake leads" as well. Automation software usually has built-in form validation techniques to eliminate bogus entries or downgrade the score based on personal email addresses. In the B2B world, most lead scoring systems are based on a numbering scale, vs A, B, C, etc. For example, using scoring up to 1000 will allow for greater differentiation of a marketers leads. This helps sales better prioritize the "hot" or "cold" leads while not being limited to choices available via a numeric lead scoring system. Great article on lead scoring and hopefully B2C and B2B lead scoring, and differences therein, can be considered in future posts.

  7. I found this article on lead scoring to be very interesting. I noticed the article is tailored towards business to consumer (B2C) lead scoring vs. business to business (B2B) lead scoring. In the B2B world, lead scores are usually calculated on a numerical scale. Numerical scoring provides more differentiation in the quality of the lead, which helps sales people prioritize their follow up. Another take away from the B2B world is that most companies don't realize they could 'unlock' leads from their existing corporate websites. Experts estimate roughly 95% of website visitors are hidden and don't fill out a web form. With the proper lead scoring and tracking solutions businesses can identify hidden leads and score their visit based on custom scoring rules and continue to track that visitor from that point forward. This is one way to automate lead scoring and lead generation vs buying leads. Many of these capabilities are built into 'marketing automation' software packages.