Data-Driven Marketing Case Study: The Engine That Refuses to Guess

Salesforce describes data-driven marketing as an approach that “uses information from a variety of sources to gain insights into consumer behavior, preferences, and trends.” Coursera puts it more directly: it “uses customer behavior data to predict future actions.”

Both definitions assume the data is right. In practice it often isn’t: ad data goes stale, leads arrive in the CRM without the campaign that produced them, and cost data sits in systems no dashboard reads. A tool that still produces a confident number in those conditions isn’t data-driven marketing. It’s a guess with decimal places.

CDAI was built against that failure. Every night, before it makes a single decision, it checks whether the data is fresh, complete, and consistent. If it isn’t, no decisions go out.

Before CDAI makes a decision, it checks the data. These are the situations it is built to catch:

Why that’s the right answer: a Scale decision on stale data tells a business to spend more on conditions that may no longer exist. No decision is better than a confident wrong one. And without knowing which campaign produced each lead, there’s no honest way to calculate cost per lead by campaign, so CDAI shows the gap instead of guessing. Our guide to offline conversion tracking covers how to connect leads to campaigns.

CDAI was run on 9 simulated businesses across 7 industries, so the right answer was known for every decision. Each one was graded against what actually happened to that campaign’s margin.

Beyond the simulation, CDAI has been tested on a live pilot client and on our founder’s own past ad account history. The accuracy figures below come from the simulated validation.

Accuracy by decision type, from the September 2026 validation

* Cut: every miss came from one deliberately extreme stress-test business.

Quarantine detection was correct in 12 of 12 cases. It scores 0 of 12 because outcome scoring needs a person to act on the call. The math matched 100% when every dollar was checked two ways.

The same safeguards run every night on every connected account: CDAI rechecks its own math and stops on bad data instead of making decisions from it. When the data is good, the decisions are graded later against what actually happened. See the validation report for the full results.

1. Collect. Pull spend, leads, and closed revenue from the tools you already use.
2. Check. Confirm the data is fresh, complete, and consistent before trusting it.
3. Calculate. Work out real profit per campaign after every cost, not just ad spend.
4. Decide. Make one clear call per campaign (scale, hold, cut, or pause) with a confidence level.
5. Grade. Once results are in, check whether each decision was right. Our 30-day retest explains how.

Step 2 is the step this case study is about.

• Tag every lead with its campaign. Keep campaign or UTM fields on the lead record in your CRM, not just in the ad platform.
• Connect closed revenue. Decisions need to know which leads became customers. See our guide to lead quality.
• Keep ad data current. Nightly syncs beat monthly exports.
• Add the costs outside the ad account: partner payouts, refunds, chargebacks, and compliance. Our guide to marketing costs lists them.