Metadata is the foundation of a digitally-driven organization. Good data and analytics depend on solid metadata. Executional agility depends on solid metadata. Yet few organizations manage metadata comprehensively. They act as if they can improvise their way forward, without understanding how all the pieces fit together. Organizational silos think about content and information in different ways, and are unable to trace the impact of content on organizational performance, or fully influence that performance through content. They need metadata that connects all their activities to achieve maximum benefit.
Babel in the Office
Let’s imagine an organization that sells a kitchen gadget.
The copywriter is concerned with how to attract interest from key groups. She thinks about the audience in terms of personas, and constructs messages around tasks and topics of interest to these people.
The product manager is concerned with how different customer segments might react to different combinations of features. She also tracks the features and price points of competitors.
The data analyst pours over shipment data of product stock keeping units (SKU) to see which ZIP codes buy the most, and which ones return the product most often.
Each of these people supports the sales process. Each, however, thinks about the customer in a different way. And each defines the product differently as well. They lack a shared vocabulary for exchanging insights.
A System-generated Problem
The different ways of considering metadata are often embedded in the various IT systems of an organization. Systems are supposed to support people. Sometimes they trap people instead. How an organization implements metadata too often reveals how bad systems create suboptimal outcomes.
Organizations generate content and data to support a growing range of purposes. Data is everywhere, but understanding is stove-piped. Insights based on metadata are not easy to access.
We can broadly group the kinds of content that audiences encounter into three main areas: media, data, and service information.
Media includes articles, videos and graphics designed to attract and retain customers and encourage behaviors such as sharing, sign-ups, inquiries, and purchases. Such persuasive media is typically the responsibility of marketing.
Customer-facing data and packaged information support pre- and post-sales operations. It can be diverse and will reflect the purpose of the organization. Ecommerce firms have online product catalogs. Membership organizations such as associations or professional groups provide events information relating to conferences, and may offer modular training materials to support accreditation. Financial, insurance and health maintenance organizations supply data relating to a customer’s account and activities. Product managers specify and supply this information, which it is often the core of the product.
Service-related information centers on communicating and structuring tasks, and indicating status details. Often this dimension has a big impact on the customer experience, such as when the customer is undergoing a transition such as learning how to operate something new, or resolving a problem. Customer service and IT staff structure how tasks are defined and delivered in automated and human support.
Navigating between these realms is the user. He or she is an individual with a unique set of preferences and needs. This individual seeks a seamless experience, and at times, a differentiated one that reflects specific requirements.
Numerous systems and databases supply bits of content and information to the user, and track what the user does and requests. Marketing uses content management and digital asset management systems. Product managers feed into a range of databases, such as product information systems or event management systems. Customer service staff design and maintain their own systems to support training and problem resolution, and diagnose issues. Customer Relationship Management software centralizes information about the customer to track their actions and identify cross selling and up selling opportunities. Customer experience engines can draw on external data sources to monitor and shape online behaviors.
All these systems are potential silos. They may “talk” to the other systems, but they don’t all talk in a language that all the human stakeholders can understand. The stakeholders instead need to learn the language of a specific ERP or CRM application made by SAP, Oracle or Salesforce.
Metadata is Too Important for IT to Own
Data grows organically. Business owners ask to add a field, and it gets added. Data can be rolled up and cross tabulated, but only to an extent. Different systems may have different definitions of items, and coordination relies on the matching of IDs between systems.
To their credit, IT staff can be masterful in pulling data from one system and pushing it into another. Data exchange — moving data between systems — has been the solution to de-siloing. APIs have made the task easier, as tight integration is not necessary. But just because data are exchanged, does not mean data are unified.
The answer to inconsistent descriptions of customers and content has been data warehousing. Everything gets dumped in the warehouse, and then a team sorts through the dump to try to figure out patterns. Data mining has its uses, but it is not a helpful solution for people trying to understand the relationships between users and items of content. It is often selective in what it looks at, and may be at a level of aggregation that individual employees can’t use.
Employees want visibility into the content they define and create, and know how customers are using it. They want to track how content is performing, and change content to improve performance. Unfortunately, the perspectives of data architects and data scientists are not well aligned with those of operational staff. An analyst at Gartner noted that businesses “struggle to govern properly the actual data (and its business metadata) in the core business systems.”
A Common Language to Address Common Concerns
Too much measurement today concerns vaguely defined “stuff”: page views, sessions, or short-lived campaigns.
Often people compare variants A and B without defining what precisely is different between them. If the A and B variations differ in several different properties, one doesn’t learn which aspects made the winning variant perform better. They learn which variant did better, but not what attributes of the content performed better. It’s like watching the winner horse at a race where you see which one won, but not knowing why.
A lot of A/B testing is done because good metadata isn’t in place, so variations need to be consciously planned and crafted in an experiment. If you don’t have good metadata, it is difficult to look retrospectively to see what had an impact.
In the absence of shared metadata, the impact of various elements isn’t clear. Suppose someone wanted to know how important the color of the gadget shown in a promotional video is on sales. Did featuring the kitchen gadget in the color red in a how-to promotional video increase sales compared to other colors? Do content creators know which color to feature in a video, based on past viewing stats, or past sales? Some organizations can’t answer these questions. Others can, but have to tease out the answer. That’s because the metadata of the media asset, the digital platform, and the ordering system aren’t coordinated.
Metadata lets you do some forensics: to explore relationships between things and actions. It can help with root cause analysis. Organizations are concerned with churn: customers who decide not to renew a service or membership, or stop buying a product they had purchased regularly. While it is hard to trace all the customer interactions with an organization, one can at least link different encounters together to explore relationships. For example, do the customers who leave tend to have certain characteristics? Do they rely on certain content — perhaps help or instructional content? What topics were people who leave most interested in? Is there any relationship between usage of marketing content about a topic, and subsequent usage of self-service content on that topic?
There is a growing awareness that how things are described internally within an organization need to relate to how they are encountered outside the organization. Online retailers are grabbling with how to synchronize the metadata in product information management systems with the metadata they must publish online for SEO. These areas are starting to converge, but not all organizations are ready.
Metadata’s Connecting Role
Metadata provides meaningful descriptions of elements and actions. Connecting people and content through metadata entails identifying the attributes of both the people and the content, and the relationships between them. Diverse business functions need uniform ways to describe important attributes of people and content, using a common vocabulary to indicate values.
The end goal is having a unified description that provides both a single view of the customer, and gives the customer a single unified view of the organization.
Different stakeholders need different levels of detail. These differences involve both the granularity of facets covered, and whether information is collected and provided at the instance level or in aggregation. One stakeholder wants to know about general patterns relating to a specific facet of content or type of user. Another stakeholder wants precise metrics about a broad category of content or user. Brands need to establish a mapping between the interests of different stakeholders to allow a common basis to trace information.
Much business metadata is item-centric. Customers and products have IDs, which form the basis of what is tracked operationally. Meanwhile, much content is described rather than ID’d. These descriptions may not map directly to operational business metadata. Operational business classifications such as product lines and sales and distribution territories don’t align with content description categories involving lifestyle-oriented product descriptions and personas. Content metadata sometimes describes high level concepts that are absent in business metadata, which are typically focused on concrete properties.
The internal language an enterprise uses to describe things doesn’t match the external language of users. We can see how terminology and focus differs in the diagram below.
Not only do the terminologies not match, the descriptors often address different realms. Audience-centric descriptions are often associated with outside sources such as user generated content, social media interactions, and external research. Business centric metadata, in contrast, reflects information captured on forms, or is based on internal implicit behavioral data.
Brands need a unified taxonomy that the entire business can use. They need to become more audience-centric in how they think about and describe people and products. Consider the style of products. Some people might choose products based on how they look: after they buy one modern-style stainless product, they are more inclined to buy an unrelated product that also happens to have the same modern stainless style because they seem to go together in their home. While some marketing copy and imagery might feature these items together, they aren’t associated in the business systems, since they represent different product categories. From the perspective of sales data, any follow-on sales appear as statistical anomalies, rather than as opportune cross-selling. The business doesn’t track products according to style in any detail, which limits its ability to curate how to feature products in marketing content.
The gap between the businesses’ definition of the customer, and the audience’s self-definition can be even wider. Firms have solid data about what a customer has done, but may not manage information relating to people’s preferences. Admittedly it is difficult to know precisely the preferences of individuals in detail, but there are opportunities to infer them. By considering content as an expression of individual preferences and values, one can infer some preferences of individuals based on the content they look at. For example, for people who look at information on the environmental impact of the product, how likely are they to buy the product compared with people who don’t view this content?
Steps toward a Common Language
Weaving together different descriptions is not a simple task. I will suggest four approaches that can help to connect metadata across different business functions.
First, the entire business should use the same descriptive vocabulary wherever possible. Mutual understanding increases the less jargon is used. If business units need to use precise, technical terminology that isn’t audience friendly, then a synonym list can provide a one-to-one mapping of terms. Avoid having different parties talk in different ways about things that are related and similar, but not identical. Saying something is “kind of close” to something else doesn’t help people connect different domains of content easily.
Second, one should cross-map different levels of detail of concern to various business units. Copywriters would be overwhelmed having to think about 30 customer segments, though that number might be right for various marketing analysis purposes. One should map the 30 segments to the six personas the copywriter relies on. Figure out how to roll up items into larger conceptual categories, or break down things into subcategories according to different metadata properties.
Third, identify crosscutting metadata topics that aren’t the primary attributes of products and people, but can play a role in the interaction between them. These might be secondary attributes such as the finish of a product, or more intangible attributes such as environmental friendliness. Think about themes that connect unrelated products, or values that people have that products might embody. Too few businesses think about the possibility that unrelated things might share common properties that connect them.
Fourth, brands should try to capture and reflect the audience-centric perspective as much as possible in their metadata. One probably doesn’t have explicit data on whether someone enjoys preparing elaborate meals in the kitchen, but there could be scattered indications relating to this. People might view pages about fancy or quick recipes — the metadata about the content combined with viewing behavior provides a signal of audience interest. Visitors might post questions about a product suggesting concern about the complexity of a device — which indicate perceptions audiences have about things discussed in content, and suggest additional content and metadata to offer. Behavioral data can combine with metadata to provide another layer of metadata. These kinds of approaches are used in recommender systems for users, but could be adapted to provide recommendations to brands about how to change content.
An Ambitious Possibility
Metadata is a connective tissue in an organization, describing items of content, as well as products and people in contexts not related to content. As important as metadata is for content, it will not realize its full potential until content metadata is connected to and consistent with metadata used elsewhere in the organization. Achieving such harmonization represents a huge challenge, but it will become more compelling as organizations seek to understand how content impacts their overall performance.