Lead scoring is a Customer Relationship Management (CRM) process that involves segmenting CRM contacts based on their likelihood to make a purchase. Lead scoring is applied to both existing customers and potential customers (known as leads or prospects) to help sales reps prioritise their work and better understand the volume of sales that are likely to convert within a period.
The lead scoring methodology typically requires collaboration between the marketing and sales teams to contact potential customers, understand which ones have interest in the company’s products or services, and then pass these “warm leads” over to the sales team to follow-up, if they haven’t already converted online. This can range from crude score-based systems to complex machine learning models.
Since most B2B e-commerce businesses are undergoing a process of digital transformation, the majority want customers to transact via the website, rather than passing them all over to the sales team to handle, since the cost to serve each customer is significantly lower online, and there are better opportunities to scale the business profitably through this channel. However, this does depend on the customer requirements.
In B2B e-commerce the sales process tends to be much longer than in B2C and customers are often more expensive to acquire and service. During the sales cycle, customers commonly require human contact to help them understand the product, negotiate the right price or configuration, arrange delivery to their facility, or provide quotations for pricing that may not be disclosed to general e-commerce website visitors.
These aspects, and particularly the quotation process, make handling B2B e-commerce orders much more time-consuming and expensive than the transactional relationship achieved through the typical e-commerce site. By implementing lead scoring to rank prospects, a sales team can separate low quality leads from those that are “sales ready”, thus making it easier for them to sell.
To work properly, a lead scoring program requires the collaboration of both sales and marketing departments to target potential prospects with marketing campaigns and encourage them to complete “micro-conversions”, such as visiting the website, filling in contact forms, viewing high value pages such as a pricing page, or downloading white papers. When put together, these actions increase the resulting score assigned to the prospect, and hopefully also increase their propensity to purchase.
Most common CRM platforms, such as Salesforce and Hubspot, now include a lead scoring system to help sales teams identify sales ready leads. However, the underlying lead scoring software within these CRM platforms work in different ways and has varying levels of sophistication. Here’s a quick summary of the main types of lead scoring model:
Manual lead scoring, or rule-based lead scoring, is the most commonly used method and is found in CRM platforms such as Salesforce. This rather crude technique requires the sales and marketing teams to identify data points stored in the CRM and assign numeric point values to each one, resulting in each prospect receiving a point value based on the various marketing and sales touch-points they have completed, which the creator thinks are linked to making a purchase.
By devising their own scoring criteria, marketing and sales teams are responsible for using their customer knowledge to help identify qualified leads and assign a lead score. A higher lead score should identify customers further through the buying cycle, or further down the funnel, who should have a higher conversion rate when contacted with sales calls or marketing messages. Here’s an example of a typical lead scoring matrix which outlines the lead score assigned for each customer action or customer segment.
|Unsubscribed from email||-2|
|Contact form submission||+25|
|Registered for webinar||+3|
|Visited landing page||+2|
|Job title: Marketing Manager||+5|
|Visited trade show booth||+3|
|Visited pricing page||+10|
Based on this lead scoring matrix, the lead scoring software would examine the CRM data held for each prospect and calculate their lead score based on their customer segmentation data or the actions they performed online or in person.
For example, here’s the data for a prospect who received a marketing email, opened it, clicked the link, browsed the site, viewed the landing page, and checked out the product pricing. They get a combined lead score of 16, which might be considered warm enough to pass over to sales if they didn’t convert online.
|Visited landing page||+2|
|Visited pricing page||+10|
As any data scientist will know, it requires strong statistical analysis skills to accurately assign lead scores because you need to understand the correlation coefficients associated with each customer segment or behavioural action, as well as their interrelationships.
It is perhaps somewhat unsurprising then to find that some authors have questioned the quality of the lead scoring results that these rule-based scoring models provide, or the ability of non-statisticians to predict the correlations that power these lead scoring tools (Marion, 2017; Bohil, 2017.)
For example, Marion (2017) examined 800 leads scored using a manual lead scoring matrix and found no statistical difference between being able to convert scored leads determined “ready for sales” and randomly selected leads, or leads with no scoring data present at all. Marion said: “there is absolutely no way that someone without experience of statistics could score or weight these activities properly”.
Bohil (2017) came to the same conclusion, so if you’re using this approach to prioritize leads or assist with your sales and marketing efforts, it would be worth getting a data scientist to clarify whether your scoring criteria actually even work.
The Ideal Customer Profile (ICP) or ideal buyer profile approach is used in the lead scoring model is in CRM platforms such as Hubspot. The ICP describes the perfect customer for an organisation and is usually based on a fictitious company that the company thinks represents the ideal. It’s particularly suited to the Account Based Marketing or ABM approach, and allows a sales team to focus on accounts that are a good fit for the company’s target customer.
Hubspot says: “If done correctly, an ICP can help define the problems you’re solving for, align your product/service capabilities with customers’ needs, and assist in laying out your future road map for product/service updates and changes.” However, in practice, the ICP approach shares many of the drawbacks that manual lead scoring has.
Unless you’ve analysed the correlations of the various customer segments associated with your “best” customers, do you really have the technical ability to pull it off correctly. It would be a tough job for a data scientist to pull off based on educated guesswork, even after examining the data, so the chances of a sales or marketing person being able to do it with a decent level of accuracy are debatable.
The final lead scoring approach - predicting lead scoring - aims to solve the issues that underlie manual leading scoring and ICP. It is the most modern and sophisticated approach - mainly because it’s based on actual data science, rather than the educated guessing of your sales team.
Predictive lead scoring systems use a machine learning classification algorithm to predict the perceived value of leads based on historical data, either predicting whether a lead will convert or won’t convert, or by predicting a probability of conversion.
This technique can use a whole range of explicit data and implicit data points to form model features, can learn from historic data, and unlike the other two approaches, is easily measurable allowing you to confirm its accuracy. Since it doesn’t require a human to calculate a lead score, these models can utilise a wider range of B2B segmentation data and look at their interrelationships, rather than just single variables.
At a computational level, predictive lead scoring models are in the same family as other models used for predicting conversion. For example, sales response models and purchase intention models both work in a similar way, analysing customer segmentation variables and behavioural data to predict the probability of a user converting.
Irrespective of the technique you apply to generate lead scores, the marketing and sales processes that utilise these data are generally fairly similar. The actual approach used partly depends on the size of the business, and the amount of resource available. Some sales teams want hot leads handed to them on a plate, others want to create the relationships and nurture leads themselves. Here’s the basic approach:
Since all lead scoring models require the use of customer segmentation data, a crucial first step is to segment the prospects to provide the data from which the lead scores can be produced. Not all segments will have predictive value in identifying which leads are more likely to convert, so a customer segmentation analysis should be undertaken.
This data science process helps to identify which segments matter to conversion probability. Once this is identified, the marketing and sales team can ensure these segmentation variables are collected from propspects, or purchased from data resellers to make the process work correctly.
Next, the marketing team will deliver targeted marketing to the prospects with the aim of getting them to perform an action that might indicate they have an interest in purchasing. This could range from traditional catalogue marketing to email marketing activity.
Customer engagement is then measured through opens, clicks, and on-site engagement goals, such as viewing a product or pricing page, or requesting a quote or further information. The data on these Marketing Qualified Leads (or MQLs) may be collected in a variety of platforms, but usually make their way to the CRM where they can be used to calculate lead scores.
Now that the initial list of prospects have been identified and have been targeted via marketing campaigns, and their behavioural and other data have been collected, the data can be utilised by the lead scoring system. This will assess each of the prospects, examine their behavioural and other segmentation variables, and assign a lead score.
If the score is low, the customer would remain in the marketing funnel and perhaps go into a separate “lead nurturing funnel” to try and increase their conversion probability. If the score is high, or if the model predicts a high rate of conversion probability, these Marketing Qualified Leads become Sales Qualified Leads (or SQLs) and get passed to the sales team to follow-up.
Finally, the sales team would work through a list of Sales Qualified Leads and make contact with them. At this final stage, these SQLs have received marketing on the company’s products or services, have expressed some kind of interest, and may have even shown some signs of being ready to purchase. At this stage, the sales team would contact the SQLs to see if they can generate a sale or conversion.
In larger businesses, the marketing team may be responsible for identifying prospects, turning them into MQLs, and then vetting them before passing them, as SQLs, to the sales team via a process called Sales Qualification. However, in small companies, the sales team would do the legwork themselves, including the identification of the potential prospects to whom they want to market.
Matt Clarke, Thursday, August 05, 2021