PHASE 1:
CREATE A 360-DEGREE VIEW OF CUSTOMER DATA
Every customer touchpoint produces valuable customer data, but with legacy data architectures, purchase data, website traffic data, email and mobile app data, paid media data, loyalty program data, and data from other categories might be stored in different places. This makes it onerous for data scientists to build attribution models that draw from a variety of sources, and the siloed data makes it nearly impossible to create personalization models that rely on real-time information. Different business units tend to see data filtered through different dashboards, giving only a partial view of the larger picture.
Meanwhile, the rapid growth of the marketing technology stack over the last decade adds to this complexity. Tools that help marketers perform web analytics, build ecommerce platforms, play videos, optimize emails, and more, also produce valuable data that often remains siloed.
To have a true 360-degree view of customers, marketers need access to both structured and unstructured data. (Structured data is what’s put into specific fields, such as what a salesperson might enter into Salesforce, while unstructured data is free-form, such as the content of customer reviews.) They also need a single source of truth in the form of clean, merged data sets that a variety of teams can use.
Finally, it’s crucial that this 360-degree view of the customer does not depend on engineering support. Advanced analytics require agility and speed, so business users need to be able to add data sources on their own without having to request engineering resources.
Ingesting from Data Sources
To avoid burdening engineers with ongoing data maintenance, which can be costly and slow, marketing organizations need an ETL tool with prebuilt connectors to data sources. These solutions extract data from the original source (for example, Facebook or Adobe), clean it or change it into a useful form for the company’s purposes (such as converting full addresses into zip codes), and load it into their data warehouse.
ETL tools worth exploring include Fivetran, Segment, and Alteryx. When vetting ETL vendors, marketing organizations should consider what percentage of their data sources are included as prebuilt connectors, and whether users can add new sources without engineering help.
Storing the unified data
Companies need a single platform like Snowflake that can natively support semi-structured data (for example, JSON data from a website) and structured data in the same system. They also need an elastically scalable platform that enables large numbers of users to run a variety of concurrent workloads from personalization to attribution to ad hoc analysis. Scalability ensures that an organization has the compute resources to introduce advanced analytics without slowing down other processes.
Accessing and querying the data
Many enterprises are starting to recognize that analytics must be available to users beyond a handful of data scientists and data analysts for data to drive results. Marketing organizations need a BI analytics platform such as Tableau or ThoughtSpot that can provide self-serve analytics to most team members.
Generally, an “80-20” rule applies to making analytics self-service for non-technical users. If they can quickly access the data they need on their own 80% of the time, the effort has been successful. (The other 20% of the time, they may need help from an analyst to write a complex ad-hoc query.)
To achieve an “80-20” balance, analytics teams must closely collaborate with business users to understand which operational dashboards and data sets they need and identify how to make requests self-service through dashboards, pre-canned queries, or pivot tables. Some companies have made tremendous strides in this direction. One public company provides a Snowflake account to every person in a business unit, for example.
Unfortunately, some BI solutions that call themselves “self-service” are not meant for business users with only general, basic data skills. Those kinds of incomplete solutions should be avoided. Business users need to quickly find information on their own and understand how campaigns are performing in real time with specific audience segments without enlisting an analyst to help. If the process to access data is onerous, business users will be unable to incorporate data into decision-making, leading to missed revenue and inefficient marketing spend.