Digital innovation has ushered in many exciting opportunities for marketers, but managing the data associated with that innovation can usher in some anxiety. Trying to explore data productively requires resources, especially in organizations where a mix of legacy tech and new solutions exists. But by incorporating dataops – developer-based protocols for data access, blending, and transformation – marketers can better define what data management efforts will raise campaign quality and ROI.
For a time, website data from web analytics was all that marketers needs to manage. The data presented limited opportunities for manipulation, and marketers simply used results from analytics solutions to determe outcomes and support decisions.
Today’s analytic capabilities cover greater complexity, meant to address more dimensions and metrics from more data sources in an organization. They also call for applying systematic means to deliver value. Agile content delivery, for example, creates metadata that must be understood in order to manage end items well. For content delivery, it means organizing tagged media for convenient query.
Being systematic raises the need to understand the data associated with the system. That need gives rise to protocols such as exploratory data analysis (EDA). EDA identifies where data is called into a key function in an advanced model such as a neural network or supervised machine learning.
Because of the high volume of data created, the risk of poor decision-making based on that data has significant consequences; problems can be existential, far more than the risks businesses encountered in the early web analytics days. Marketing campaigns can be delayed, machine learning models can be infused with poor assumptions and negative bias, and technological hiccups exposed to a customer experience can irrevocably sabotage brand image.
Enter dataops, a set of concepts and categories that help understand and manage data, related operations, and the supporting elements. Dataops can offer teams guidance for blending unstructured data against needed operations, and for managing overall analysis time. Typically dataops is used for advanced data models, but as devices, and subsequently operations, become more data-driven, the concepts from dataops can be applied more broadly.
Thus using dataops can create a data-influenced standard operating procedure for marketers that should enhance productivity. Communication on metrics and analytic solutions can better match marketing needs, because a shared understanding is derived from the data. This can lead to better decisions on tasks associated with content and analytics.
Dataops also evaluates the origins of data beyond its current storage, be it a data lake or database. A marketer can first ask if data is sourced from the following frameworks:
- Transactional Data – metrics associated with a purchase or sale, such as when a person buys, how the purchase was made, and the application of a discount, if any
- Interactional Data – metrics from media such as websites, apps, social media, phone, email
- External Data – data from outside an organization that can augment a model.
Marketers can then discern which operational practices and solutions intersect with the given data types and storage capabilities to improve business requirement parameters, data quality, and data security. A deep technical expertise is not required to map out these practices. Marketers inexperienced with database technology can ask some good starting questions such as:
- How should alerts be qualified, noting what degree of data changes indicates an outlier — by a percentage or by a threshold value??
- What kind of data should be stored temporarily, such as using a data lake and an extended ETL (extract, transform, load — a solution that stores data in any format until it can be used in an operation or model)?
- How should tables be arranged to fulfill the desired business requirements? ?What columns will each table contain? ?
- How much data will be populated into these tables? ?
- What kind of webhooks and APIs will provide the data? ?
These considerations lead to a key benefit from a dataops protocol — the elimination of poor communication among teams that leads to costly data errors. One group of users may vet data differently than another group does because they have different criteria as to what constitutes quality data.
Poor communication regarding how data is labeled or accepted can also introduce bias and error in programmatic initiatives that support digital marketing campaigns. Good ongoing communication can keep data safe by providing transparency into the data quality supporting an algorithm.
Furthermore, dataops can enhance self-service analytics by surfacing key processes that impact the organization. The data preparation in a self-service analysis can involve time-consuming tasks such as gathering requirements, modeling data, creating a report, and arrange for distribution of reports. Self-service analytics can lead to siloed analytics, but at best it incorporates solutions that allow a broad number of people to share and review results. Dataops processes can take individual self-service activity to the next level by linking those tasks to the organizational schedule, aligning the activity with project deadlines and schedule alterations.
Dataops practices hold the potential promise to align the goals of the different data users (including marketers), improving the focus of data science, business intelligence, operations, and IT.
The end result should be a dynamic improvement in data value, in lockstep with a dynamic business environment.