Computer-implemented adaptive subscription models for consumer packaged goods

A non-transitory computer-readable storage medium comprising computer program instructions that when executed by one or more processors effectuate operations comprising:

• obtaining, in a database, a plurality of user records, each record comprising household properties having corresponding values and consumption amounts for goods;
• forming, from the plurality of user records, a training set and a validation set;
• training, based on the training set, a consumption model configured to output, based on input values for a set of a plurality of household properties represented in the training records, determined consumption amounts for respective goods represented in the training records, wherein the model minimizes error of output determined consumption amounts relative to the consumption amounts for goods in at least some of the user records in the validation set;
• obtaining a new user record comprising at least some household properties having corresponding values;
• determining, with the consumption model based on the new user record, a predicted consumption amount for each good in a set of goods;
• obtaining, for a good in the set of goods, at least one stock record indicative of at least one stock keeping unit (SKU) having a count of the good that satisfies in whole or in part the predicted consumption amount for the good, wherein at least one SKU is obtained for each good in the set of goods and a combination of SKUs is obtained for at least some of the goods in the set of goods;
• generating, for the set of goods, an order comprising a plurality of SKUs indicated in obtained stock records for which a total count of one or more SKUs corresponding to respective ones of the goods meet the predicted consumption amount for the good without exceeding a threshold;
• and
• submitting the generated order for fulfillment.