Systems and methods for real-time lead grading

A method comprising:

• receiving historical lead data, wherein the historical lead data comprises a plurality of leads and a plurality of conversion event records for the plurality of leads, and wherein each lead of the plurality of leads comprises a plurality of lead attribute values;
• converting the plurality of lead attribute values of each of the plurality of leads into a corresponding plurality of benchmark values based on the plurality of conversion event records;
• determining a first conversion rate for the plurality of leads based on the plurality of conversion event records;
• determining a second conversion rate for a subset of the plurality of leads based on the plurality of conversion event records, wherein the subset comprises leads of the plurality of leads with a first lead attribute value;
• determining a benchmark value by dividing the second conversion rate by the first conversion rate;
• and
• replacing the first lead attribute value by the benchmark value in each lead of the subset of the plurality of leads;
• encoding the plurality of leads into a plurality of lead vectors;
• clustering the plurality of lead vectors to form a plurality of clusters;
• determining a plurality of cluster labels for the plurality of clusters based on the plurality of conversion event records;
• and
• determining a mapping from a lead vector space to the plurality of cluster labels by:
• pairing a lead vector of the plurality of lead vectors with a cluster label of a cluster of which the lead vector is a member, to form a training data pair;
• and
• training a neural network to learn the mapping from the lead vector space to the plurality of cluster labels using the training data pair.