click-through rate prediction
In plain wordsEstimating how likely a particular person is to click on a particular ad, search result or recommended item: a prediction made billions of times a day to decide what each user is shown.
The task of estimating the probability that a user will click an item (an advertisement, search result, product or post) if it is shown in a given context, from features of the user, the item, the query and the situation. In online advertising sold per click, ads are typically ranked by bid times predicted click-through rate (pCTR), an estimate of expected revenue per impression, so predictions must be calibratedcalibration, n.: How well a model's stated confidence matches reality: when a well-calibrated system says it is 80% sure, it turns out to be right about 80% of the time. as well as correctly ordered; similar scores drive recommenderrecommender system, n.: Software that guesses which songs, products, posts or jobs you are most likely to want and shows those first, as when a video app lines up the next clip or a shop suggests an add-on. ranking. The data are vast, very sparse and imbalanced, with clicks often around one percent of impressions or fewer, and dominated by high-cardinality categorical features whose interactions matter. Model families evolved from logistic regressionlogistic regression, n.: A simple, widely used method for yes-or-no predictions that outputs a probability — say, a 70% chance a customer cancels — by passing a weighted sum of inputs through an S-shaped curve. on hashed feature crosses trained by online learningonline learning, n.: Learning from data that arrive one piece at a time, updating the model after each new example instead of training once on a fixed collection, as a spam filter does while emails stream in., through boosted trees feeding linear models and factorization machines, to deep models that combine learned embeddingsembedding, n.: A list of numbers that stands for a word, picture or other item, arranged so that similar items get similar lists. It lets a computer measure similarity of meaning as distance. with explicit feature-interaction layers, attention over user histories and transformer-based generative recommendationgenerative recommendation, n.: A newer approach to recommendation that uses language-model-style AI to predict the next item a person will want, one step at a time, much as a chatbot predicts the next word.. Typical metrics are AUCAUC, abbr.: A single number from 0 to 1 summarizing how well a model ranks real positive cases above negative ones: 0.5 is no better than a coin flip and 1.0 is perfect., log losslog loss, n.: A score for probability predictions that punishes confident mistakes severely: saying '99% sure' and being wrong costs far more than saying '60% sure' and being wrong. Lower is better. and calibration.
A 0.1 percent gain in offline AUC for the CTR prediction model moved revenue measurably in the A/B test.
Origin
Click-through rate became an advertising metric with mid-1990s web banners; predicting it became central once search ads were priced per click and ranked partly by expected clicks, as in Google's AdWords from 2002. Matthew Richardson, Ewa Dominowska and Robert Ragno's 2007 paper on predicting clicks for new ads is an early academic treatment; papers from Google (2013) and Facebook (2014) described production systems.
Usage note
Optimizing for clicks can reward sensational or misleading content (clickbait), so platforms blend click prediction with objectives such as dwell time, conversions and quality signals. Logged clicks reflect what was shown rather than what users would have chosen, a source of position and selection bias.
Related
- recommender systemrecommender system, n.: Software that guesses which songs, products, posts or jobs you are most likely to want and shows those first, as when a video app lines up the next clip or a shop suggests an add-on.
- learning to ranklearning to rank, n.: Training a model to put items in the best order — search results, products, recommendations — by learning from examples of which items people judged more relevant or chose more often.
- logistic regressionlogistic regression, n.: A simple, widely used method for yes-or-no predictions that outputs a probability — say, a 70% chance a customer cancels — by passing a weighted sum of inputs through an S-shaped curve.
- online learningonline learning, n.: Learning from data that arrive one piece at a time, updating the model after each new example instead of training once on a fixed collection, as a spam filter does while emails stream in.
- two-tower modeltwo-tower model, n.: A design for recommendation and search in which one neural network turns users or queries into lists of numbers and another does the same for items, so good matches can be found by quick comparison.
- generative recommendationgenerative recommendation, n.: A newer approach to recommendation that uses language-model-style AI to predict the next item a person will want, one step at a time, much as a chatbot predicts the next word.
- calibrationcalibration, n.: How well a model's stated confidence matches reality: when a well-calibrated system says it is 80% sure, it turns out to be right about 80% of the time.
- uplift modelinguplift modeling, n.: Predicting not who will buy, recover or respond, but whose behavior will actually change because of an action such as a discount or a treatment, so that it is offered only where it makes a difference.
Sources
- Richardson, M., Dominowska, E. & Ragno, R. (2007). Predicting Clicks: Estimating the Click-Through Rate for New Ads. Proc. WWW 2007.
- McMahan, H. B. et al. (2013). Ad Click Prediction: a View from the Trenches. Proc. KDD 2013.
- He, X. et al. (2014). Practical Lessons from Predicting Clicks on Ads at Facebook. Proc. ADKDD 2014.
Mentioned in
uplift modelinguplift modeling, n.: Predicting not who will buy, recover or respond, but whose behavior will actually change because of an action such as a discount or a treatment, so that it is offered only where it makes a difference.