Lead Scores

Goal

Businesses want to enhance the effectiveness of their marketing and sales resources. This means finding customers or potential customers that can be moved down the purchase funnel with greater efficiency. This means increasing the marginal utility of every dollar spent on campaigns or sales efforts. Lead scores represent a targeting tool that enables focusing energy and resources on higher yield opportunities.

Lead Score is a prediction of how likely an individual is to make a purchase on a scale of 1-100, where higher numbers represent a higher likelihood to purchase. This score operationalizes a Bayesian framework wherein a sliding degree of information about an individual contributes to a lessening uncertainty in their likelihood to transact with a vendor. The degree of uncertainty can be assessed in terms of time intervals that can be tuned to business models and typical vertical conventions (e.g., high repeat CPG versus infrequent B2B; or CPG vs financial services

Types of Lead Scores by Data Source

  • e-commerce using website engagement [links to LCV and product recommender with website touches as proxies for transactions]
  • simple transaction [links to LCV]
  • no transaction [links to contact lists and marketing engagement]

Primary Source Data This is a custom model per company’s sales transactions. Training data needs to include 5000 users with and without purchase history (1/2 of each is ideal). With regular retraining of your company’s transactions, this model will always be up-to-date. The time to retrain should be every few months. It largely depends on purchase behavior (and non-purchase behavior) of your customers. Retraining each month is acceptable, given enough new data or new products available. Include the full transaction history for each person (up to two years).

  • transactions
  • demographics and lifecycle
  • product or brand match scores
  • engagement indicators e.g., marketing interactions, ad touches, website visits

Minimum Data Requirements for model building

Note that transaction and engagement history can be substitutions

  • PII
  • purchase or transaction history
  • engagement history (marketing outreach or website interaction)
  • brand or product match scores

Minimum Data Requirements for no transaction/engagement for contacts

  • PII
  • demographics
  • list qualifications

Deliverable

  • record-level scores 0-10 for lead
  • Lead Score = likelihood to purchase with specified time period (i.e., range 0-10, where a larger number corresponds to a greater likelihood of purchase). Recommended time periods are: 30 days, 60 days, and 90 days; but likelihood estimates are highly dependent on data sources and vertical constraints.
  • Included with time interval lead scores is a corresponding uncertainty coefficient functions as a validation metric. The uncertainty coefficient has three classes: green, yellow, red. This classification indicates the quality and predictive power of the data associated with the lead score. Lead scores can be grouped by colour bands to indicate confidence in scores or data quality issues that needs to resolved at the data source.

Primary Source Data

  • Minimum is binary outcome behaviour (transactions, purchases, marketing engagement, etc.)
  • Record-level profiles/bios or related personal descriptions
  • Brand/Product Statements
Share the Post:

Related Posts

Wrench.AI partners with Ideaship to scale personalization

Lead Scores

Goal Businesses want to enhance the effectiveness of their marketing and sales resources. This means finding customers or potential customers

Read More