MiQ is a leading global programmatic media partner helping agencies and brands plan, execute and optimize campaigns across a fragmented advertising ecosystem. With massive data volumes exceeding 700 trillion fragmented signals across multiple internal systems, MiQ needed to move beyond siloed platforms, manual processes and unreliable pipelines to achieve its ultimate vision - a media planning agent that could reason across this immense data volume in natural language. The company set out to build a connected intelligence layer that could unify audience discovery, market research, campaign planning and optimization into a single experience. MiQ built Sigma on the Databricks Platform to unify governed data, analytics and AI into a single experience. With Genie’s conversational interface, teams can now use AI agents to plan campaigns, generate recommendations and make confident decisions in minutes.
Fragmented data ingestion introduced risk in client campaigns
MiQ’s vision is to lead the programmatic industry and make it better. A key element to this is building AI-powered intelligence that, gives marketers and their agencies a platform- and data-agnostic way to turn customer engagement signals into better campaign strategy and execution. To bring that vision to life, MiQ built Sigma, their award-winning advertising platform to serve as a single pane of glass for campaign planning, trading, analytics and experimentation. Sigma helps internal teams and clients reason across 600+ data feeds, 1.7 billion audience profiles, historical campaign performance and real-time pacing signals. But the goal was broader than analytics; MiQ also sought to develop a Planning Agent capable of helping users leverage natural language interactions to optimize campaign planning with richer, more tailored planning insights in a fraction of the time of manual planning.
But MiQ’s existing environment made that level of intelligence difficult to scale. “There were too many data silos. Teams worked across separate internal platforms for traders, sellers and analytics, each with its own pipelines and data stores,” explained Abhishek Pandey, Global Product Lead, Intelligence at MiQ. Data from external providers flowed into a shared landing zone, then moved through cleaning, anonymization and structuring processes spread across different clusters and tools. Governance was fragmented across Hive Metastore workspaces. This made privacy, lineage, access control and reusable data products harder to manage consistently. At the same time, media planners faced a growing challenge: information existed across disconnected systems, taxonomies were difficult to harmonize, and valuable consumer insights often lacked the context needed to drive action. Planning omnichannel campaigns required teams to navigate fragmented datasets, evolving media channels and varying levels of programmatic expertise.
As Sigma became generally available to clients and internal teams, MiQ could no longer tolerate unpredictable risk to customer-facing experiences. To support natural language analytics, agentic planning workflows and SLA-backed campaign intelligence, Sigma required a unified data foundation built for reliability, governance and production-scale AI.
Databricks powers the governed intelligence layer behind Sigma
MiQ built Sigma on the Databricks Platform to unify their data ecosystem of 700 trillion data signals spanning watching, browsing and buying activity. External data from dozens of adtech providers lands in S3 cloud storage via StreamSets, NiFi, Kinesis and vendor APIs. The data moves through a medallion architecture on Delta Lake so teams can work from consistent, governed data products across activation, analytics and agentic workflows. Unity Catalog provides the governance layer behind this foundation, helping MiQ manage fine-grained access control, lineage and reusable data assets across teams. Roughly 1,000 ETL jobs are managed via Declarative Automation Bundles (formerly known as Declarative Automation Bundles) for continuous integration and delivery (CI/CD). Databricks SQL powers the serving layer for dashboards, Genie Spaces and agent queries, giving Sigma a scalable way to support both user-facing analytics and production AI experiences without adding operational burden.
On top of that foundation, MiQ uses Genie and Databricks-managed MCP Servers to give agents access to governed business data. Sigma’s Planning Agent uses two Genie spaces, one for unified consumer signals and one for historical campaign data, to reason across audience behavior, campaign performance and media planning questions. Vector Search helps turn natural language audience descriptions like “people who enjoy baking with chocolate in New York” into actionable activation segments, while foundation model integrations support things like ranking and persona generation. With the Planning Agent, non-technical marketers can use natural language to go from audience discovery to media plan recommendations in a matter of minutes rather than days, drastically reducing planning cycles. The Planning Agent acts as a unified destination for agency planners and MiQ commercial teams during critical moments such as pitching, responding to briefs and identifying upsell opportunities. Users can explore audience segments, analyze competitors, uncover market trends and plan channel allocations through a single conversational interface. According to Pandey, “Without Databricks, this agent wouldn’t be possible. With Genie, our teams can easily interact with complex campaign and audience data to make decisions that improve our clients’ ad performance.”
Building the Planning Agent required more than connecting an LLM to enterprise data. MiQ had to orchestrate reasoning across diverse datasets, including consumer intelligence, internal knowledge hubs and live campaign performance data. The team developed an agentic execution framework capable of maintaining context across sessions, minimizing latency and producing more consistent, trustworthy outputs. To improve accuracy, MiQ combined automated evaluation techniques with human feedback loops to establish a gold standard for campaign recommendations and planning outcomes.
This architecture also supports MiQ’s Trading Agent, which gives traders a chat interface for campaign insights and recommendations. By connecting agentic workflows to the same governed data foundation that powers dashboards, pipelines and activation, MiQ can make Sigma more than a front-end experience. It becomes a production intelligence layer for planning, trading and optimizing media with greater consistency, trust and control.
Sigma accelerates smarter campaign decisions and better customer experience at global scale
MiQ has since achieved significant efficiency and productivity gains with Databricks. Campaign planning and analysis that once took 6+ hours can now happen in roughly 5 minutes through agent-assisted workflows. Work that previously required teams to manually gather information across research tools, audience datasets and campaign systems can now happen through a single conversational experience. Planners and traders have moved from manual investigation to faster, more consistent decision-making.
The business impact is already immediate across campaign performance. Sigma has powered over 40,000 campaigns for more than 2,300 advertisers since launch and won multiple technology awards. MiQ’s use of Genie-powered agents has enabled double the campaign optimizations leading to better results - testing shows that Sigma campaigns return up to $2.22 in value for every $1 spent when compared to standard programmatic campaign setups. By bringing governed data and AI into day-to-day campaign workflows, teams can detect underperforming campaigns faster, surface explainable recommendations and optimize media strategies with greater confidence.
MiQ has also strengthened the operational foundation behind Sigma. Databricks SQL improved serving performance from 22 seconds to 4 seconds, while Declarative Automation Bundles helped reduce release cycles from roughly 3 weeks to about 4 hours. AI-assisted optimization workflows reduced storage by 50%, saved $85K per year and improved query performance by 25%, helping MiQ manage petabyte-scale Delta estates more efficiently.
Next, MiQ plans to expand its use of Databricks across more agentic and operational workflows. The team is migrating more ETL and streaming jobs to serverless jobs after an initial cohort showed 45% average cost reduction and 60% performance improvement. MiQ also plans to advance its Trading Agent from insight generation to read/write capabilities, so traders can adjust budgets, tactics and pacing directly from the agent while maintaining governance and auditability through Databricks. The long-term vision is to make Sigma the primary destination for planning and optimizing campaigns, creating a connected intelligence platform that helps agencies and brands move from fragmented information to actionable media decisions at global scale.
Watch MIQ's talk at Data + AI Summit 2026 to learn more.