Abstract: When customizing a sales prediction model for a particular client, there exists a cold-start problem when the client's training data is of low quality and/or quantity. Disclosed embodiments solve this cold-start problem by normalizing the data based on weighted or unweighted recency and/or trend. In addition, the data from a wide range of clients may be used to train a base sales prediction model, which can then be fine-tuned using a particular client's data to produce a client-specific sales prediction model. A self-monitoring mechanism may be used to continually retrain the client-specific sales prediction model on new client data, to gradually improve upon the accuracy of the client-specific sales prediction model as more data become available.

Abstract: The diversity of job titles prevents the extraction of information from job titles to be automated and scalable. Accordingly, disclosed embodiments utilize a machine-learning model to classify job titles by one or more characteristics, such as job level and/or job function. The characteristic(s) may be extracted from the job titles to be used as an input to a persona model that predicts a persona score, indicating the relative importance of a person to a sales opportunity.

Abstract: The diversity of job titles prevents the extraction of information from job titles to be automated and scalable. Accordingly, disclosed embodiments utilize a machine-learning named-entity recognition model to tokenize job titles, and tag those tokens according to one or more named entities, such as job responsibility and job function. The named entities may be extracted from the tagged tokens to be used as an input to a persona model that predicts a persona score, indicating the relative importance of a person to a sales opportunity. This enables granular information to be automatically extracted from job titles, regardless of structure, style, or format, without continual retraining and in a scalable manner, for use in one or more downstream functions.

Abstract: The determination of an account outcome, such as a renewal, upsell, or cross-sell, for a product subscription is not presently capable of automation or scaling. Accordingly, disclosed embodiments automate this determination, in a scalable manner, using a trained predictive machine-learning model. In particular, values for a set of features of a customer account are extracted and input to a predictive machine-learning model to identify the probability of an account outcome. This probability may then be used in one or more downstream functions, such as for reports, alerts, account segmentation, marketing orchestration, sales intelligence, determining a next best action, and/or the like.

Abstract: The Internet generally provides anonymity to the online activities of visitors to web sites and other online resources. This prevents the operators of web sites and others from identifying visitors who do not wish to be identified. Accordingly, embodiments generate mappings between entities (e.g., IP addresses, domains, cookies, or devices) and accounts (e.g., companies) to de-anonymize online activities. In an embodiment, summary mappings are generated based on activity data. Each summary mapping may comprise an entity, potential account identifier, and an activity vector that measures observations of an association between the entity and potential account identifier from an activity source for multiple summary periods. A model may be applied to the summary mappings to compute signal strengths for a plurality of candidate mappings. A winning mapping may then be selected for each entity in the candidate mappings, and used to associate the entity with an account in one or more downstream functions.

Abstract: AI-based orchestration. In an embodiment, a recommendation engine is applied to a data pipeline representing company accounts. Engagement metric(s) are calculated based on activity data associated with the company accounts, and predictive model(s) are applied to the activity data and/or firmographic data associated with the company accounts to generate predictive output. A tactic recommendation model is applied to orchestration features, comprising the engagement metric(s) and predictive output, to generate recommended tactic(s). In addition, a contact recommendation model is applied to contact data to generate recommended contact(s). The recommended tactic(s) are combined with the recommend contact(s) to generate an orchestration, comprising recommended action(s), to be executed.

Abstract: Artificial intelligence for keyword recommendations. In an embodiment, raw keyword data are received. The raw keyword data comprise keyword activity records that each comprises a uniform resource locator (URL) for an online resource and metadata for the online resource. Arrays of keywords are extracted from the keyword activity records, with each array of keywords associated with the URL in the keyword activity record from which the array of keywords was extracted. User-specified keyword(s) are received, and a subset of the arrays of keywords that match at least one of the user-specified keyword(s) is identified. A training dataset is generated from the subset, and used to train a machine-learning model to output recommended keywords based on an input keyword.