1. Knowledge Base
  2. Solution Guide
  3. Machine Learning & Data Science Concepts

Data for Machine Learning

Review a use case of how Narrative can provide data for machine learning purposes all from one marketplace.

Problem

DataAnalytics is building a new platform that is designed to provide in-depth analysis of mobile consumer behavior - where & when they shop, buying habits, digital consumption behavior, and more. The project’s main blocker is finding the best way to source the data they need to feed their models. They continuously run into issues of standardization, cleanliness, and duplication of the data they’ve purchased in the past when sourcing  from multiple vendors. 

They have turned to Narrative to be their one-stop-shop data marketplace. The first use they would like to explore is buying data to help build brand affinity scores across groups of MAIDs. 

  • Buy various data sources based on different habits & activities 
  • Understand frequency of observations - times visited/purchased/opened
  • Control deduplication strategies of data points 

 

Narrative Solution

Through Narrative, DA can identify data that they believe are key indicators to building brand affinity scores. 

  • MAIDs who have visited brand locations, based on frequency of visits
  • Branded purchase history habits - items, item category, merchant, frequency of purchases
  • MAIDs with brand loyalty apps on their phone, measure frequency of use & amount of in-app purchases

 

Summary

Narrative works with standard schemas across all available data packages and does not believe in creating pre-packaged segments. This allows buyers to purchase from different packages - from location data to branded purchase history - combine them to a single user ID source, and make their own inference to what this data may say. 

Controlling the deduplication strategy gives buyers the option to determine how often they want to buy the same data point. For DataAnalytics, buying the same MAID makes sense if they want to understand how often a single MAID visits a store or how often that MAID makes in-app purchases every month. This can help build indexes, as a MAID who has visited a store 5x in the past 30 days displays a higher brand affinity than a MAID who has visited just once. 

By combining data points like visits + purchase history + loyalty app use, as well as understanding varying levels of frequency of each data point (times visited, time in app), a company like DataAnalytics can begin to paint a clear picture of a user's level of affinity to specific brands. 

Want to learn more about how Narrative can help you take control of your data strategy? Contact partner-success@narrative.io today.