The Plato Glossary

Lead generation, bidding, attribution and direct mail terms, each defined in one sentence and linked to the article that explains it.

40 terms

A

Agentic AI: Agentic AI is software built on large language models that plans and takes multi-step actions toward a goal using tools, such as reading contracts or monitoring buyers, while fast numeric decisions like bidding stay with classical models.Read: Agentic AI in Lead Gen: Where LLMs Help, Where They Don't
Attribution window: An attribution window is the lookback period after a click or impression during which a later conversion is credited to it, and the right length is where a channel's cumulative conversion-lag curve flattens.Read: Attribution Windows: Finding the Optimal Lookback Period

B

Bid shading: Bid shading is bidding below your true valuation in a first-price auction, at the price that maximizes win probability times profit, because there you pay exactly what you bid.Read: First-Price vs. Second-Price Auctions: Strategic Lead Buying
Budget pacing: Budget pacing adjusts bids throughout the day, often every few minutes, so a lead or media budget is spread over the whole day instead of being exhausted early or left unspent.Read: Real-time Bid Optimization for Affiliate Publishers

C

Contextual bandit: A contextual bandit is a multi-armed bandit that conditions each choice on features of the user or session, so it learns the best offer for each segment rather than a single overall winner.Read: Multi-Armed Bandits: Beyond A/B Testing for Affiliates

D

Data enrichment: Data enrichment is appending third-party data to a lead or household record, and it pays for itself only when the extra information is likely to change a decision worth more than the data costs.Read: Data Enrichment: When Third-Party Signals Pay for Themselves
Dynamic pricing: Dynamic lead pricing values each lead individually in real time from its predicted conversion probability and value to each buyer, replacing fixed price tiers that overpay for weak leads and underprice strong ones.Read: The End of Fixed Bidding: Dynamic Pricing for Lead Gen

E

EPC (earnings per click): EPC, or earnings per click, is the revenue an offer or placement earns divided by the clicks it receives, and it is the standard yardstick for ranking affiliate offers and offerwall slots.Read: The Offerwall Engine: Ordering Offers for Maximum EPC
Expected value of information: The expected value of information is how much better a decision becomes with an extra piece of data multiplied by what that decision is worth, which tells you whether buying the data is worthwhile.Read: Data Enrichment: When Third-Party Signals Pay for Themselves

F

First-price auction: In a first-price auction the highest bidder wins and pays exactly what it bid, so bidding true value leaves no profit and buyers must shade their bids.Read: First-Price vs. Second-Price Auctions: Strategic Lead Buying

G

Generative engine optimization (GEO): Generative engine optimization (GEO) is the practice of making your content the source that AI answer engines such as ChatGPT, Perplexity and Google AI Overviews quote and recommend.Read: GEO for Lenders: How to Get Recommended by AI Answers

H

Heuristic rules: Heuristic rules are hand-written if-then policies, such as "if credit is above 600, send to Lender X", that treat everything near a threshold as a cliff and never update themselves from data.Read: From Heuristics to Algorithms: Moving Beyond 'If-Then' Rules
Holdout group: A holdout group is a randomly chosen set of prospects or traffic that deliberately does not receive a treatment, so comparing it with the treated group measures the treatment's true incremental effect.Read: Optimizing the Cadence: How Often to Send Direct Mail

I

Incrementality: Incrementality is the share of conversions that happened only because of a campaign, channel or partner, as opposed to conversions that would have happened anyway.Read: Attribution Windows: Finding the Optimal Lookback Period
Intent model: An intent model predicts how ready a borrower is to apply right now from session behavior such as entry query, pace, depth and hesitation, complementing a credit score that only describes the past.Read: Predicting Intent: The Digital Body Language of Borrowers

L

Last-click attribution: Last-click attribution gives all the credit for a conversion to the final touchpoint, which overfunds branded search and starves the channels that start lead journeys.Read: Attribution Models for Multi-Touch Lead Journeys
Lead fraud: Lead fraud is the sale of fake or misrepresented leads, from bots and fraud farms to synthetic identities and recycled leads, and machine learning models catch it by scoring behavioral, device, network and velocity signals in milliseconds.Read: Real-Time Lead Verification: Reducing Fraud with ML
Lead returns: Lead returns are leads a buyer sends back to the seller for credit under agreed rules, such as invalid contact details, duplicates or fraud, often weeks after the sale.Read: Real-Time Lead Verification: Reducing Fraud with ML
Lookalike modeling: Lookalike modeling trains a model on past responders and non-responders to score new prospects by how likely they are to behave like your best customers, not merely to share their demographics.Read: Lookalike Modeling for High-Value Direct Mail Prospects
LTV bidding: LTV bidding sets bids from a lead's or click's predicted lifetime value, such as funded amount, repeat borrowing and losses, rather than from the first conversion alone.Read: Real-time Bid Optimization for Affiliate Publishers

M

Mail cadence: Mail cadence is how often and at what intervals a household receives direct mail, best set per household by predicting when response peaks and when fatigue sets in.Read: Optimizing the Cadence: How Often to Send Direct Mail
Model drift: Model drift is the decline in a model's accuracy as the data it sees or the outcomes it predicts shift away from what it was trained on, such as when lender appetite, rates or fraud patterns change.Read: Build vs. Buy vs. Partner: AI in the Foundation-Model Era
Multi-armed bandit: A multi-armed bandit is a testing algorithm that shifts traffic toward better-performing options as evidence accumulates, losing far less revenue during a test than a fixed A/B split.Read: Multi-Armed Bandits: Beyond A/B Testing for Affiliates
Multi-touch attribution: Multi-touch attribution splits credit for a conversion across all the touchpoints in a journey, ideally in proportion to each one's measured incremental contribution.Read: Attribution Models for Multi-Touch Lead Journeys

O

Offerwall: An offerwall is a page that shows a visitor a ranked list of offers, and it earns the most when offers are ordered by expected revenue for that visitor (payout times predicted click-through times predicted conversion) rather than by headline payout.Read: The Offerwall Engine: Ordering Offers for Maximum EPC

P

Ping-post: Ping-post is a two-step lead sale in which the seller first "pings" buyers with partial, anonymized lead data to collect bids, then "posts" the full lead, contact details included, to the winning buyer.Read: Latency is Revenue: The High-Frequency Trading of Lead Gen
Prescreened offer: A prescreened offer is a firm offer of credit mailed to consumers selected from credit bureau data under the permissible-purpose rules of the Fair Credit Reporting Act.Read: Lookalike Modeling for High-Value Direct Mail Prospects
Price elasticity: Price elasticity is how strongly demand responds to a change in price, such as how much a lead buyer's purchase volume falls when the lead price rises.Read: The End of Fixed Bidding: Dynamic Pricing for Lead Gen
Proxy conversion: A proxy conversion is an early signal, observed within minutes of a visit, that is statistically correlated with a downstream outcome that arrives days later, so bidding models can learn without waiting for revenue.Read: Real-time Bid Optimization for Affiliate Publishers

R

Reserve price: A reserve price is the minimum bid a seller will accept in an auction, protecting the value of leads when competition is thin.Read: First-Price vs. Second-Price Auctions: Strategic Lead Buying

S

Second-price auction: In a second-price auction the highest bidder wins but pays roughly the runner-up's bid, so bidding your true value is the dominant strategy.Read: First-Price vs. Second-Price Auctions: Strategic Lead Buying
Shapley value: A Shapley value, from cooperative game theory, credits each participant, such as a marketing touchpoint, with its average marginal contribution across every possible combination of participants.Read: Attribution Models for Multi-Touch Lead Journeys
Smart routing: Smart routing ranks buyers and offers for each lead by expected value (price times probability of conversion) rather than price alone, and optimizes the whole sequence a user sees.Read: Smart Routing: Sequencing Lead Offers for Maximum Yield
Speed-to-lead: Speed-to-lead is the time between a consumer submitting a form and the first contact from the buyer, and it is one of the strongest drivers of whether a lead converts.Read: The ROI of Real-Time Decisioning in Lead Marketplaces

T

Thompson sampling: Thompson sampling is a bandit strategy that picks each option with the probability that it is the best, by sampling from a posterior estimate of every option's performance.Read: Multi-Armed Bandits: Beyond A/B Testing for Affiliates
Trigger lead: A trigger lead is a consumer flagged by a credit bureau shortly after a credit inquiry, signaling that they are actively shopping for a loan.Read: Lookalike Modeling for High-Value Direct Mail Prospects

U

Uplift modeling: Uplift modeling predicts how much a treatment, such as a mail piece, changes each person's probability of responding, so you target the persuadable rather than people who would have responded anyway.Read: Optimizing the Cadence: How Often to Send Direct Mail

V

Velocity signals: Velocity signals count how often the same device, IP address, email or identity appears within a short time, and sudden spikes are among the clearest markers of lead fraud.Read: Real-Time Lead Verification: Reducing Fraud with ML

W

Waterfall routing: Waterfall routing offers each lead to buyers one at a time in a fixed priority order, usually by price, which ignores each buyer's probability of converting the lead.Read: Smart Routing: Sequencing Lead Offers for Maximum Yield
Winner's curse: The winner's curse is the tendency of the winning bidder to overpay when an item's value is uncertain, because winning means its estimate was the most optimistic.Read: The Winner's Curse: Avoiding Overpayment in Lead Auctions