Lead scoring model – why most fail

And how is not this the most reprehensible ignorance, to think that one knows what one does not know? But I, colleagues! in this, perhaps, differ from most marketers; and if I should say that I am in any thing wiser than another, it would be in this, that not having a competent knowledge of the lead scores of our leads, I also think that I have not such knowledge. -Socrates, Marketer

So you have a marketing automation system, and you’re implementing a multi-model system covering implicit and explicit data. Your sales team is going to spend less time chasing cold leads, they’ll know everything about their prospects, and your revenue is about to triple.

Not so fast.

If your lead scoring model is like most, it will fail in the worst possible way: sales will see great leads with low scores, bad leads with high scores, and conclude “this score means nothing.”

Looking Deeper

To understand this, lets consider how most models work.  Jon Fortune searches for widgets and clicks on your ad(+5), registers on your landing page for a whitepaper (+5), and downloads the whitepaper from the email you send him (+5).  You ask about company industry and he’s in the right industry (+5).

So far, you’ve collected demographic information(industry) and behavior information(actions on website) – good work!  But your threshold for marketing qualified leads is 30, so Jon will remain in your marketing nurturing queue for awhile, hopefully consuming your content and eventually raising his hand and asking to talk to sales.

Heres the problem: Jon Fortune is the CEO of a construction company in Nevada.  Construction isn’t your top niche, but your sales guy in Nevada (Mark) prefers construction leads, because he has experience in the industry and closes them easily.  Being a busy CEO, Jon doesn’t use a lot of email (he has an assistant for this) so he doesn’t open your nurturing emails.  He has a pressing need for your widgets, the authority and budget to purchase, and wants to buy soon.  While sitting in your nurturing cycle, your competitor cold calls and reaches him, setting an appointment and eventually gets the deal.  Mark learns of this deal through acquaintances, searches the database and finds out Jon was in marketing being nurtured.  Mark then threatens to quit unless he gets all leads in his territory, without qualification.

What happened here?  Your scoring model assumed it knew everything, so it didn’t qualify the lead properly.  Perhaps 70 percent of the leads in construction are scored correctly, but the system didn’t know Mark is great at closing these leads.  Result: Sales loses confidence in nurturing, thinks the scores are wrong, and your marketing automation investment is considered a failure.

Another scenario: Jane visits you at a trade show and enters her business card for your tschotske(+10).  Jane has no authority or budget to buy your widgets (she was just laid off), but she desperately wants to work for your company.  So she goes to your twitter and follows you (+10!), likes your Facebook page (+10!!), and visits 20 pages on your website to learn more(+20!!!).  She applied through an external website so she never visits your careers page, but in a short span she now has 50 points, enough to label her a marketing qualified lead.

This time, Mark calls Jane, excited to get a qualified lead from Marketing.  The number he has from her business card goes to an office voicemail, but he doesn’t despair: this one is clearly interested.  He calls back, ten times, and never receives a response.  Finally, he gets a response from her former boss:  ”Jane doesn’t work here anymore, and we have no need for your widgets.”  Rejected.

This time, your scoring model failed because it was too sensitive, quickly judging a lead to be qualified based on actions that didn’t matter.  It doesn’t ask the question “is Jane still employed?” because its too difficult to answer.  Instead, it makes assumptions and hopes for the best.

Score the Prospect Upside and Win

The first step is to mentally acknowledge the problem and account for it: understand the limits of lead scoring, and acknowledge these to sales.

The next step is to make the limits explicit.  Each lead should have a score based on known demographics, but it should also have a “potential score”, indicating the maximum demographic score the lead could ever have.  In contrast to the behavioral and demographic scores, this potential score will actually decrease as you learn more.  The difference between the demographic score and the potential score is the “prospect upside” and this should be a calculated field, visible to sales and marketing.

Now that you report the data for the problem, you can minimize it and measure the improvement.  For example, you could look at a quarter of data and see the average “prospect upside” is 50 points, and make it a goal to decrease this to 20 points.  You could then look at the fields with the biggest impact and purchase the information.  Using Salesforce?   Jigsaw would fix this.

While you’re working to improve the data, you can educate sales on the different scores.  They intuitively know a lead with 5 points could be more valuable than a lead with 50 points, so giving them these new data points will improve their confidence in you.  Rather than holding back leads from sales, consider passing all of them over immediately, while only showing qualified leads in the default views.

Bottom Line

Marketing automation systems are quick to tell you data, and slow to tell you what data is missing.  As a result, sales can lose confidence in marketing if their expectations of scoring aren’t met.  By admitting your level of confidence in the data, you can improve their trust in marketing, and over time measure the improvement in the quality of your data.

Tagged as: lead generation scoring, lead scoring model