From Intention to Action: The How-to Guide for Data Governance in AI-Ready Commerce

Ask brands why they need data governance frameworks, and you’ll get a host of answers.

Compliance and security top the list, and analytics make the cut.

AI, meanwhile, is the up-and-comer: Organizations know that AI is quickly becoming a must-have for ecommerce operations, and that data governance plays a key role in this shift.

Research firm PwC puts it simply: “Strong data governance is a powerful enabler for the use of AI. It builds trust, simplifies compliance, and confirms that AI systems are reliable, transparent, and fair.”

“Why” isn’t enough on its own, however. Brands also need to know “how” — how do they build effective data governance policies? How do they implement them? And how do they operationalize these policies to drive digital growth?

You need the how; here’s the how-to.

4 Pillars That Define the ‘What’ of Modern Data Governance Strategies

Basics are the foundation of data governance. Get them right, and you’re set up for success in managing, using, and securing data. Trade these starting points for speed, however, and you’re facing regulatory, revenue, and reputational challenges.

Here are the four pillars of modern data governance strategies.

Pillar 1: Data Quality

Low-quality data and bad data hygiene lead to low-quality governance. If data sources are inaccurate, contain duplicate entries, or lack the standardization necessary for collection and comparison, it becomes impossible for organizations to create a single source of truth.

Delivering consistent data quality starts with validation and cleansing processes that ensure information entered is timely, relevant, and accurate. Ensuring continued data quality requires monitoring and evaluation — the sooner teams discover data issues, the better.

Pillar 2: Data Management

Data management speaks to the collection, storage, and use of data. Effective data management spans the life cycle of information, ensuring that it’s protected at each stage and is only accessible to those with the correct permissions and access.

Pillar 3: Data Security

Collected data isn’t inherently secure, which creates challenges for data governance. Improving security starts with access. Who can view, share, and edit data sources? Why? And for how long?

Other security best practices include multi-factor authentication, data encryption, and behavioral analytics to detect unexpected or insecure actions.

Pillar 4: Data Stewardship

The last pillar of data governance is stewardship. This is the assignment of accountability and responsibility. As a general rule, anyone who uses or changes brand data has some level of accountability for its safety.

Most organizations, however, also assign larger-scale responsibilities to C-suite leaders such as chief information security officers (CISOs), chief information officers (CIOs), or chief data officers (CDOs).

How To Build Better Data Governance

The four pillars of data governance are exactly that — supports for more in-depth governance strategies. With these pillars in place, ecommerce brands have the “why” and the “what” covered but aren’t sure what comes next.

How do these strategies map to sales, marketing, security, and compliance operations?

The Intent: Streamlining Token Usage

AI in ecommerce depends on a combination of content and context to produce accurate outputs. Content is the data itself; context is how this data applies in specific situations. To help manage the massive data volumes collected by AI, this data is broken down into smaller units, known as tokens.

These tokens are the building blocks of relevant and accurate AI outputs. When users ask a question, intelligent tools search for applicable tokens and use them to create answers.

As noted by Anthropic, however, this creates an engineering problem: “Optimizing the utility of tokens against the inherent constraints of LLMs to consistently achieve a desired outcome.”

In other words, while all tokens have value, not all tokens are valuable at the same time. Using sub-optimal tokens can lead to inefficient outputs and open the door to data governance issues.

The Action: Building a ‘Library of Context’

Brands can enhance token retrieval and application by building what’s known as a library of context. This is a centralized, trusted repository of token data that helps AI streamline token usage.

Common context library components include FAQs, operational guides, audience personas, and product specifications. When AI tools receive a query, they search the library for applicable tokens and use these tokens as the framework for a more in-depth answer.

In effect, the library points tools in the right direction, ensuring they don’t waste resources on irrelevant data or inadvertently expose secure information.

To build a library of context, start by analyzing common queries and responses to determine which tokens belong on the shelves.

Next, brands must ensure they have sufficient infrastructure to support a centralized, queryable knowledge base. Finally, this knowledge base must be regularly evaluated and updated to ensure it remains relevant.

The Intent: Improving Data Structure

Data governance isn’t just about security. It’s also about an operational strategy that effectively manages data so that generative AI engines — such as those now powering popular search platforms — can find and recommend your products to prospective customers.

If your data structure is full of inaccurate or duplicate entries, AI reasoning engines won’t see it as an authoritative source. Instead, the best your brands can hope for is a supporting role in general search queries.

The Action: Making a Clean Sweep

Improving data structure means putting in the time and effort required for full cleaning.

This starts with standardization — the more standardized, the better. Date formats offer an easy example. If some files are stored using the MM-DD-YYYY format and some are DD-MM-YYYY, this can cause challenges for auditability and compliance.

Deduplication and outliers are next. Redundant data can cause issues with data set analysis, leading to over- or under-reporting that impacts operations. Outliers have a similar effect. If data points are well above or below median values, they can significantly skew averages.

Deduplication is typically accomplished using automated tools that identify and remove data duplicates. Outlier analysis, meanwhile, is a human function. This is because outliers aren’t always wrong; instead, they may be leading trend indicators that require further analysis. Data management professionals can help determine if outliers are mistakes or offer meaningful insight.

Clean data offers a dual benefit. First, it improves customers' trust. AI engines are more likely to recommend clean data, and customers visiting your ecommerce site are more likely to return if product data is consistent and accurate.

Second is mastering the agentic product shelf. This is now the first shelf many customers see and is served up after every AI-enabled query. If users don’t like what they see on agentic shelves, they won’t bother with digital storefronts and are far less likely to visit brick-and-mortar stores.

The Intent: Enhancing Information Exchange

For brands to reach consumers, information must flow. More data in more places means more chances to spark interest and drive conversion.

Data governance plays a key role in this information exchange, specifically through the use of the Global Data Synchronization Network (GDSN).

The Action: Syndicating Product Data

A not-for-profit organization known as GS1 provides global standards for product identification and communication. It does so using the GDSN, which is an interconnected network of data pools.

To leverage the GDSN, however, companies need to transform and store their product data such that it meets GS1 requirements. Data covered by GS1 includes logistics, shipping, and regulatory information. Brands must also create and assign Global Location Numbers (GLNs) and Global Trade Item Numbers (GTINs) to all products.

Salsify is a notable GDSN Data Pool provider and helps companies connect their supply chain content to any trading partner in Salsify-supported markets, regardless of what data pool they use. Combined with the Salsify Product Experience Management (PXM) platform, companies can streamline GDSN data governance by centralizing and syndicating product data.

Delivering on Data Governance

AI-ready ecommerce is no longer the exception. Consumers now expect AI-enabled websites, mobile applications, and chatbots that both help them find what they’re looking for and suggest related products they didn’t know they needed.

Delivering on the promise of intelligent ecommerce, however, requires comprehensive data governance policies. While it’s possible to offer entry-level AI operations with minimal governance, systems will rapidly reach their limit once brands begin collecting and using first- and zero-party data to inform product recommendations and improve product detail pages (PDPs).

To deliver on data governance, ecommerce brands need the “how” behind the “why”. By building a library of context, making a clean sweep, and syndicating product data, brands are better prepared to act on data governance expectations.