A CMO walks into a quarterly business review wanting to know which campaigns drove revenue last quarter. The question is reasonable. A week later, the team is still arguing about which definition of “revenue” to use, whether the attribution model should be first-touch or data-driven, and why the figures in the email platform do not match the figures in the warehouse.
Nothing has gone wrong, exactly. The CMO got what most CMOs get when they ask a marketing question that touches data: a meeting. Then another meeting. Then a spreadsheet with three tabs and a note at the bottom that reads “directional only, subject to model assumptions”. The campaign question becomes an attribution question, the attribution question becomes a definitions question, and the definitions question becomes a procurement question for a tool that promises to fix all of it.
It rarely does. The five questions every senior marketer asks reduce to the same underlying problem, and that problem is not a tooling problem. Tooling assumes the substrate is already there. Most of the time it is not.
The question that started this
Watch enough of these meetings and a pattern shows up. The CMO asks a question. The team narrows the question to make it answerable. The narrowed version is no longer the question the CMO asked. Everyone signs off because it is Friday.
The original question was “which campaigns drove revenue”. The answerable version was “which campaign IDs appear in the last-touch position for closed-won deals in the CRM, joined to media spend in the ad platforms, for the subset of customers whose IDs survive both joins”. Those are different questions. The second one is much smaller, much less useful, and the only one the data can actually support.
This is what marketing leaders mean when they say their dashboards do not tell them anything. The dashboards are technically correct. They are answering a smaller question than the one that was asked. The gap between the two questions is where most of a marketing budget hides.
The five questions every senior marketer asks
There are roughly five questions that drive most senior marketing conversations.
- Who are our most valuable customers, and what do they have in common?
- Which channels and campaigns actually drove revenue last quarter?
- Which customers are about to churn, and which are about to expand?
- Where is our acquisition cost rising fastest, and why?
- What happens to a customer between the moment they enter our marketing stack and the moment they appear in the finance system?
Every one of these is, on its surface, a marketing question. Every one of them, in practice, is a data question. Each one needs a single durable answer to: who is the customer, what did they do, when did they do it, how do we know it was them, and which of those facts are we still allowed to use.
A team that can answer those five base questions, cleanly and repeatedly, can answer most senior marketing questions in under a week. A team that cannot will keep buying tools, hiring analysts, and running discovery sessions, and the meetings will keep ending with “directional only”. The next four posts in this series each pick up one of those five questions and look at the data shape required to actually answer it.
Why MarTech tools alone fall short
The MarTech category is not short of capable products. The problem is what they assume coming in. A campaign management tool assumes the audience definitions are correct. A customer data platform assumes the inputs are clean, deduplicated, and consented. An attribution platform assumes the event stream is complete and the identifiers stitch. A reverse ETL tool assumes the warehouse model it is pulling from is trustworthy.
None of those assumptions are free. Each of them is a piece of engineering work. When the tool is bought before the work is done, the tool inherits whatever was wrong with the substrate, and now there is a vendor logo attached to the wrong answer.
The pattern is consistent across categories. Customer data platforms are sold on the promise of a unified customer profile, then take eighteen months to produce one because the source systems disagree about what a customer is. Attribution platforms are sold on the promise of channel clarity, then produce confident numbers based on tags that fire half the time. Personalisation engines are sold on the promise of one-to-one experiences, then ship segments of one million because the only stable identifier in the system is “all visitors”.
This is not a criticism of the vendors. The vendors are doing what they sold. The gap is the layer beneath them, and that layer is identity, consent, event capture, and a warehouse model that everyone in the business agrees with. Once that layer exists, the MarTech tools work the way the demos suggested. Without it, they do not.
What teams that get this right look like
The teams that answer marketing questions quickly tend to share a few characteristics. They treat the customer data model as a product, with an owner, a roadmap, and a definition of done. They have one definition of “customer”, one definition of “revenue”, and one place where those definitions live. They invest in event capture before they invest in activation, because no amount of downstream tooling can repair a thin upstream signal. They keep consent close to the data, not in a separate compliance silo, so that the question “are we allowed to use this” is answered by the same query that pulls the audience.
A useful real-world reference point is the multi-year engagement with Craveable Brands, the parent company behind Red Rooster, Oporto, and Chicken Treat. Marketing leadership there did not start with a personalisation engine. They started with a unified data foundation, a customer identity model that worked across brands, and event capture that the marketing, finance, and operations teams all agreed on. The activation layer came later, and it worked, because the substrate underneath it was sound.
Teams that get this right also pick their fights. They do not try to fix everything at once. They pick the two or three questions that matter most to the business this quarter, work backwards to the data those questions need, and build only that. The result, six months in, is fewer tools, fewer dashboards, faster decisions, and a CMO who can answer the revenue question in the meeting where it was asked.
The shape of this work, identity, consent, event capture, agreed definitions, sits in the awkward space between marketing operations and data engineering, which is exactly where Sakura’s Data & AI practice was set up to operate.