Why Age-Based Maintenance is Failing Multifamily

Ask 10 multifamily operators how they decide when to replace a water heater, and most will give you a version of the same answer: “We replace them at twelve years,” or “When the unit fails.” Either approach is a rule of thumb dressed up as a strategy. Neither one reflects what we know about how building components fail.

The cost of getting this wrong is not theoretical. The National Apartment Association’s most recent Survey of Operating Income & Expenses in Rental Apartment Communitiesshows that repairs and maintenance—together with the administrative overhead tied to managing them—remain among the largest controllable lineitems for stabilized properties. The U.S. Department of Energy has documented in multiple Building America Solution Center articles that an HVAC system running to failure costs two to three times more than one replacedon a planned schedule, before factoring in collateral damage or resident displacement.

In other words, the standard age cutoff is wrong roughly half the time. Half of the components that a portfolio replaces under the 12-year rule still had useful life remaining. The other half were already operating past the point where the math favored replacement.

What Time-Based Maintenance Optimizes For

It’s worth being honest about why age-based maintenance became standard. It’s easy. It’s an approach that requires no data beyond an install date. It survives staff turnover. And in an industry where most maintenance information lived in paper folders until the last decade, it was often the best-available option.

But what time-based scheduling optimizes for isn’t equipment reliability or CapEx efficiency. It optimizes for scheduling tractability—the ease of putting something on a calendar—which doesn’t account for the many reasons units across portfolios may benefit from maintenance on different schedules.

A nine-year-old electric water heater in Phoenix, on water that runs 18+ grains per gallon of hardness, with no anode rod replacements logged, is a different asset than a nine-year-old electric water heater in coastal Oregon on softened municipal water. The Energy Information Administration’s residential consumption data shows that hot-water demand alone varies by more than 35% across U.S. climate regions. The Insurance Institute for Business and Home Safety has published a similar variation on HVAC and roofing failure rates across climate zones, citing humidity, salt-air exposure, and freeze-thaw cycling as primary drivers. An age cutoff ignores all of it.

What “Predictive Maintenance” Really Means

The term predictive maintenance—sometimes called condition-based maintenance—has been around in industrial settings for decades. Multifamily is later to the conversation than manufacturing, aviation, and commercial HVAC, but the underlying math is identical.

At its core, predictive maintenance treats every component as a probability problem rather than a yes/no issue. Three concepts matter:

• Survival distributions. A statistical curve (the Weibull distribution is the most common) that describes the probability a component is still working at any given age. It’s the same math used in life insurance underwriting and aircraft component scheduling. The shape of the curve differs by component class—water heater anode rods, HVAC compressors, roof membranes, and electrical panel breakers all have very different failure profiles—and shifts with environment and brand quality.

• Bayesian updating. Rather than relying on age alone, this model takes in additional evidence—a tech’s note about bearing noise on a blower motor, a recurring service call for low refrigerant, an insurance claim from a sibling unit—and updates the component’s failure probability in response. The longer a unit operates without these signals, the more its risk score drops relative to a same-age peer.

• Cascade and correlation effects. Component failures don’t happen in isolation. A clogged air filter measurably raises blower-motor and compressor strain; a failed pressure-relief valve on one water heater is mild evidence that the same install batch elsewhere in the portfolio is also at risk. Modern predictive systems explicitly model these dependencies.

The output is not “this unit is old.” The output is “this component has an X% probability of failing in year Y, with a confidence range of Y–A to Y+B.” That’s the type of output that’s truly useful for a CapEx plan, an insurance underwriting conversation, or a vendor scheduling decision.

Why This Finally Works for Multifamily

Three things changed in the last few years that made predictive maintenance practical for apartment

portfolios. First, data accessibility. Property management systems, like Rent Manager, now expose properties, units, and work orders through documented APIs. The work-order history that used to be locked in written or PDF service tickets is increasingly structured data.

Second, environmental data is free and granular. NOAA’s degree-day datasets, USGS water hardness maps, and EPA air-quality data are all available at zip-code resolution for free. The inputs that determine 40–60% of the variance in component life are sitting in public databases.

Third, computing is no longer the bottleneck. Running a Weibull-Bayesian model across 30 component types and a thousand units used to require a data team. ForVue can run it for you nightly on a single server.

In practice, this means that a portfolio doesn’t need a McKinsey project to begin moving from reactive to predictive maintenance. It needs three things in this order: structured work-order history (which most operators already have inside Rent Manager), an explicit list of components per unit with install dates, and a model—purpose-built or vendor-provided—that combines those inputs with environmental and observation data.

A Practical First Step

For operators not yet doing this, the lowest-risk way to start is to pick one component class—water heaters are a good candidate because failures are expensive, frequent, and well-documented—and run a parallel forecast against your current age-based schedule for one year. Track the predictions, log the actual failures, and compare. The exercise itself will sharpen your existing data hygiene and surface gaps in your install-date records.