What Are You Actually Building Your Tech Stack On

Microsoft Fabric has won the platform conversation in multifamily analytics. The architecture is sound. The integration with the rest of the Microsoft stack is meaningful. The vision of a unified data and analytics platform is the right vision.

The thing operators are starting to discover is that picking Fabric is not the same as having Fabric.

The platform is one component of an analytics environment. The data model that runs on it is another. The pipelines that feed it are a third. The semantic layer that makes its outputs interpretable to the business is a fourth. The governance framework that keeps it honest is a fifth. Picking the platform is the first percent of the work. The other 99 percent is what builds the advantage.

The 12 to 18 month reality

Industry implementations of enterprise data platforms have followed a consistent timeline. Gartner research found that on average only 48 percent of AI projects make it into production and it takes 8 months to go from prototype to production (Gartner, via Informatica 2025). That is for the projects that succeed. The ones that do not are still in motion at 12, 14 and 18 months in.

The reasons are not mysterious. Building a multifamily-specific semantic data model takes domain expertise that most internal teams have to develop on the job. Cleaning legacy data takes longer than anyone budgets for. Governance frameworks have to be debated, written, approved and adopted. The pilot phase is followed by a production phase that is two to three times longer than the pilot.

Most of that time is invisible to the executive sponsor. The deck shows steady progress. The reality on the ground is that the team is building infrastructure that does not produce a screenshot.

The platform is the easy decision. The 18 months that follow are where most operators lose their lead.

What it actually costs

The build cost of a Fabric-based multifamily analytics environment, done internally, breaks down into four buckets that most ROI models do not capture.

Direct labor cost. A small team of data engineers, data analysts and platform architects working for 12 to 18 months. At fully loaded rates, this lands between 1.5 and 3 million dollars depending on team composition.

Opportunity cost. The decisions not being made during the build window. Revenue management running on legacy reporting for another year. Budgeting cycles that start with the same reconciliation problem as last year. AI use cases that cannot launch because the data is not ready. This cost is the largest of the four and the hardest to estimate. McKinsey’s research suggests organizations with strong data integration achieve roughly threefold higher returns from AI than those with weak data connectivity (McKinsey, via Integrate.io, 2024).

Risk cost. RAND research showed that over 80 percent of AI projects fail to reach production deployment (RAND Corporation, 2024). For an internal Fabric build, the probability that the team ships exactly what was scoped, on time and to a quality that the business will actually use, is not 100 percent. It is something less. The expected value of the project has to be adjusted accordingly.

Maintenance cost. Once built, the environment has to be maintained. Microsoft Fabric is updated continuously. Multifamily data sources evolve. The team that built the original environment becomes the team responsible for it indefinitely.

What the buy alternative looks like

REBA exists because the multifamily industry needed a different answer to the same question. Rather than every operator spending 12 to 18 months building the same foundation, we built it once for the industry and ship it as a platform.

What that means in practice. The Fabric environment is configured. The multifamily-specific semantic model is built. The data pipelines from the major property management systems are productized. The governance framework is operational. The enrichment with third-party market data is live.

REBA customers today are not starting from a blank Fabric tenant. They are starting from a multifamily data foundation that 114 other multifamily operators are already running on. The first business question can be answered within the first quarter, not the second year.

The question is not whether to build on Fabric. The question is whether to spend 18 months building the foundation yourself or to deploy one that already exists.

The honest math

For some operators, building internally is the right answer. Operators with very specific use cases that no platform can address, with strong existing data engineering teams and with a longer time horizon will sometimes get a better outcome from a custom build.

For most operators, the math does not work that way. The internal build costs more, takes longer, carries more risk and ties up scarce talent on infrastructure work rather than on the use cases that differentiate the business.

The companies pulling ahead in multifamily right now are not the ones with the largest internal data teams. They are the ones who decided where to spend their internal data team’s time. The foundation is the same work for every operator. The advantage is what each operator builds on top of it.

Sources cited

• Gartner, via Informatica. Only 48 percent of AI projects reach production, with 8 months average from prototype to production. 2025.
• RAND Corporation. Over 80 percent of AI projects fail to reach production deployment. 2024.
• McKinsey, via Integrate.io. Organizations with strong data integration achieve approximately threefold higher AI ROI than those with weak connectivity. 2024.

BUY THE FOUNDATION. BUILD YOUR ADVANTAGE.