Tips for data teams building Campaign Intelligence

As more and more brands take control of their advertising data, more data teams are being asked to help build out a single pane of glass for their organization to understand their advertising and marketing investments using the Snowflake Marketing Data Cloud. This ask is driven by several trends, including:

• The resurgence of Marketing Mix Modeling (MMM) and the data it needs to be successful.
• The desire for brands to own their own data in a single platform rather than it being in silos.
• The desire for applications (like MMM) to come to the data rather than moving data to the applications.

Data teams are often experts in data engineering, data warehousing, and platforms like Snowflake, but they may not be experts in advertising and marketing channels. They may have never worked with data from Facebook, Google Ads, Tiktok and the dozens of other sources required. This post, then, will give some practical tips for data teams building out a Campaign Intelligence solution.

1. Beware the build of connectors — It can seem simple to build connectors to pull data from marketing channels like Facebook, Tiktok, your email platform, etc. Most of these platforms provide simple APIs, so it can seem excessive to buy a SaaS tool to merely call a few APIs. Especially because some of these tools call hundreds of APIs — if you only need a dozen or so, aren’t you paying for a lot of connectors you don’t need? Not so fast. Consider that you are signing up to maintain these connectors, not just to build them. Platforms can change the API, breaking your connector, and requiring you quickly pull a developer off another project to fix it. In addition, marketing teams can quickly add a new channel to a media plan — you don’t want to be left scrambling or miss out on important data. Building can be the right answer if you have the team to support or you have security requirements that can best be met with an internal tool, but be sure to consider the full commitment.
2. Check the channels — You have done your due diligence, talking to other teams that have implemented a similar solution. You’ve heard what they like and what they don’t like about the tools they use to gather data. Before you run off and sign up for the same tool, check to make sure the tool lines up with the channels you use. You can almost ignore the largest channels — every tool will be able to get data from Facebook. It’s the smaller channels that might drive the decision — what channels does your business use that most don’t? Even further, what channels do you use that don’t even have an API? Ask your marketing team what channels send a daily or weekly email instead of having a UI or API. Ask about billboards or traditional radio — these channels might not be top of mind, but they might be the difference between a holistic solution and an incomplete one. These “hard to reach places” might determine success, so have a plan for them.
3. Deep dive on Dashboards — Like other BI-focused use cases, a lot of the initial use of this data will be to power dashboards to give marketing teams and executives insight into where their marketing dollars are going and how they are being used. To understand what queries can be run on the fly and what needs to be materialized, you’ll want to understand what dashboards will be used. How many analysts are there and how often do they look at the data? How does that compare to the usage by executives and other stakeholders? This will help you with planning for both cost and performance.
4. Master Measurement — The advertising data collected will often be used for marketing measurement as a next step. Although data engineers are not expected to be experts in marketing measurement, it can be useful to understand the different types of measurement to know how the data will be used. From Marketing Mix Modeling (MMM) to Multi-touch Attribution (MTA) to Sales lift, understanding what techniques there are will help you understand the data required.
5. Make the model — I spoke with Sumit Bhatia, Field CTO at Kipi, to get his advice, since Kipi has helped multiple customers build out their campaign intelligence solution. His advice was to focus on building out a Common Data Model (CDM) for marketing data that can handle the use cases. This helps the data engineering team in their fight for clean, consistent data. Then, as you ingest sources, they can be mapped to your CDM. This model should be well documented, serving as a semantic layer that can be used to democatize insights using LLMs.
6. Require row-level logs — The starting point for reporting on campaign data is usually aggregated data. The aggregated data can give valuable insights on impressions delivered, spend, and clicks across channels. Some use cases, however, like reach calculations or lift measurement, require person-level data. If these uses cases are, or might become, part of the roadmap, keep track of which channels make row-level data available. Think about techniques you will use to ingest that data and have an architecture that can handle the scale.
7. Honor History — Switching to a new data platform for marketing data can be of limited value if you only load data going forward. Many use cases are more powerful with historical data. Understanding how metrics change over time can put valuable context on the analytics. Marketing Mix Modeling requires historical data (at least a year — two is better) to train the model. So as you plan the project, make sure you also have a plan to load sufficient historical data.
8. Don’t forget the future — While planning the project, you will of course discuss with the marketing team their requirements for the project. And while you never want to overengineer a system, it could be beneficial to have some sense of where the project might go. It could help you build in some flexibility for where things might go and how priorities and requirements might change.

Marketing is becoming increasingly data driven, and a partnership with data teams is essential for them to have the data they need in a scalable, flexible platform. We have seen many large advertisers build campaign intelligence systems on Snowflake, so we hope these tips can help data teams work with marketing teams to build a system that will satisfy current requirements and stand the test of time.

Ready to get started? Check out the Marketing Data Foundation Quickstart.