{"id":156,"date":"2026-03-24T15:40:43","date_gmt":"2026-03-24T15:40:43","guid":{"rendered":"https:\/\/thelab.wppresolve.com\/?p=156"},"modified":"2026-04-28T09:57:43","modified_gmt":"2026-04-28T09:57:43","slug":"data-quality-assurance-agent-blog-post","status":"publish","type":"post","link":"https:\/\/cms.research.wpp.com\/?p=156","title":{"rendered":"Meet Your New Agentic Data Guardian"},"content":{"rendered":"\n

<\/p>\n\n\n\n

1. The high cost of \u201cdirty\u201d data<\/strong><\/h2>\n\n\n\n

In the modern enterprise, data is the raw ingredient behind every strategic decision. Think of it like a premier restaurant: the Data Engineer is the sous-chef, meticulously sourcing and preparing ingredients, while the Data Scientist is the executive chef, transforming them into the predictive models and insights that drive the business forward. If the ingredients are spoiled or mislabelled, the final dish fails, no matter how talented the chef.<\/p>\n\n\n\n

Across several of our AI initiatives at WPP, we uncovered a pattern that was quietly draining velocity from our most ambitious projects. Our \u201csous-chefs\u201d, skilled data engineers responsible for pipeline integrity, were spending up to one full day per week<\/strong> on tedious, largely manual Quality Assurance (QA) of data flowing into BigQuery. Row by row, column by column, they checked for missing values, logical contradictions, and phantom duplicates, work that was essential but deeply repetitive.<\/p>\n\n\n\n

This wasn\u2019t just an inconvenience. It was a strategic bottleneck<\/strong>: it slowed the delivery of every downstream AI application, consumed senior engineering talent on janitorial tasks, and most dangerously created risk. When a human eye is the only safeguard between raw data and a production model, errors don\u2019t just slip through occasionally. They slip through systematically<\/em>, at exactly the moments when the data is most complex and the engineer is most fatigued.<\/p>\n\n\n\n

We asked ourselves a different question: What if, instead of building another dashboard or writing another validation script, we built an intelligent agent, one that could reason about data quality the way an experienced engineer does, learn from every audit it performs, and get better over time?<\/em><\/p>\n\n\n\n

This article describes how we built that agent, what makes it fundamentally different from traditional automation, and what happened when we put it to the test.<\/p>\n\n\n\n

\n\n\n\n

2. The problem: why data quality demands more than scripts<\/strong><\/h2>\n\n\n\n

The data & the modelling ecosystem<\/strong><\/h3>\n\n\n\n

The agent operates on digital marketing campaign performance data hosted in BigQuery, massive tables that track how advertising campaigns perform on a daily basis across major ad networks like Meta (Facebook and Instagram). Each row represents a highly granular intersection of a specific campaign, audience segment, platform, device, and creative asset. This data captures everything from broad identifiers (like the parent brand and geographical targeting) down to precise performance metrics, including impressions, clicks, daily spend, conversions, leads, and app installs.<\/p>\n\n\n\n

This foundational data is the lifeblood of two critical machine learning systems:<\/p>\n\n\n\n

    \n
  1. The Prediction Model:<\/strong> A classification system designed to predict whether a planned campaign will yield a negative, neutral, or positive outcome.<\/li>\n\n\n\n
  2. The Recommendation System:<\/strong> A highly flexible advisory engine capable of handling any combination of “missing modalities.” For example, if a media planner inputs a specific Brand, Target Audience, and Location, the system dynamically recommends the optimal missing parameters, such as the best platform to use and the most effective creative asset to deploy.<\/li>\n<\/ol>\n\n\n\n

    For more background on the broader modelling context, see From Guesswork to Glimpse: How AI is Predicting the Future of Marketing Campaigns<\/a>.<\/p>\n\n\n\n

    The silent threat of data corruption<\/strong><\/h3>\n\n\n\n

    Because these models directly inform real-world media spend and strategic campaign planning, their accuracy is paramount. The underlying data is regularly refreshed directly from the advertising platforms to keep the models up to date. However, this automated refresh process frequently introduces subtle corruption and systemic inconsistencies.<\/p>\n\n\n\n

    For instance, while metrics like engagement<\/code> and clicks<\/code> generally remain stable, downstream pipeline issues frequently render conversions<\/code> and awareness<\/code> metrics unreliable (“not high quality”). At the individual row level, these anomalies are often entirely invisible. But at scale, they are devastating. If left unchecked, these untrustworthy data points bleed into the training sets, silently degrading the prediction model’s accuracy and causing the recommendation engine to suggest sub-optimal, expensive campaign configurations. This makes rigorous, automated data quality validation not just a nice-to-have, but an absolute necessity for the ecosystem to function.<\/p>\n\n\n\n

    The failure modes<\/h3>\n\n\n\n

    The scale and velocity of data flowing into BigQuery mean that errors don\u2019t announce themselves. They hide. Through our manual QA process, we catalogued six prevalent failure modes, each one capable of silently degrading every model built on top of the data:<\/p>\n\n\n\n

    Failure Mode<\/th>What Happens<\/th>Why It Matters<\/th><\/tr><\/thead>
    Missing Values<\/strong><\/td>Fields arrive empty: sometimes 5% of a column, sometimes 40%<\/td>Models trained on incomplete data learn incomplete patterns. Forecasts drift silently.<\/td><\/tr>
    Outliers<\/strong><\/td>A metric reads 200,000 clicks when the true value is 500<\/td>A single extreme value can skew an entire model\u2019s calibration, distorting spend recommendations.<\/td><\/tr>
    Duplicate Rows<\/strong><\/td>Identical records appear multiple times<\/td>Inflated counts cascade into inflated budgets. Campaigns appear to outperform reality.<\/td><\/tr>
    Categorical Corruption<\/strong><\/td>A brand name like \"Nike\"<\/code> is replaced with \"zX9pQ\"<\/code><\/td>Segmentation breaks. Reports attribute performance to entities that don\u2019t exist.<\/td><\/tr>
    Logical Inconsistencies<\/strong><\/td>More clicks than impressions. Spend recorded against zero impressions.<\/td>These are the most insidious. Each value looks valid in isolation, but the relationships<\/em> between them violate business reality.<\/td><\/tr>
    Missing Columns<\/strong><\/td>An entire field disappears from a refresh<\/td>Downstream pipelines fail or, worse, silently fall back to defaults.<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

    A static validation script can catch some of these: the easy ones, the ones you’ve already seen. But scripts are brittle<\/strong>: they encode yesterday\u2019s assumptions and break on tomorrow\u2019s edge case. They cannot reason about why<\/em> a pattern looks wrong, weigh it against historical context, or decide whether a recurring anomaly is a genuine error or a known artifact of a data source.<\/p>\n\n\n\n

    That requires judgment. And judgment is what we built the agent to provide.<\/p>\n\n\n\n

    \n\n\n\n

    3. Our approach: an agent that reasons, remembers, and improves<\/h2>\n\n\n\n

    We designed the Data Quality Assurance Agent<\/strong> as a reasoning entity<\/em> capable of planning an audit strategy, querying data, forming hypotheses about its health, testing those hypotheses, and learning from the results. The distinction matters. A script checks what you tell it to check. An agent decides what to check<\/em>, based on what it knows and it has the tools to act on that decision end-to-end.<\/p>\n\n\n\n

    Architecture: one agent, specialised tools<\/h3>\n\n\n\n

    The agent is powered by a single reasoning core that plans, decides, and acts. What gives it breadth is its toolkit<\/strong>, a set of specialised capabilities it can invoke as needed, selecting the right tool for each step of the audit:<\/p>\n\n\n\n

    \"\"
    Data Agent Architecture Diagram<\/figcaption><\/figure>\n\n\n\n