How to Size a UPS by Real Watts, Runtime, and Battery Derate
The two errors that compound in UPS sizing
UPS sizing fails in two predictable ways, and they compound. The first is mathematical: the system gets sized from the VA number on the front of the box, while the real limit is the watts rating on the datasheet. A 1500 VA UPS may only support 900 to 1,000 W depending on power factor — put it on an 1,100 W real load and it is already overloaded even though the VA number looked safe. The relationship is straightforward: watts equals VA times power factor. A 1500 VA unit at 0.6 power factor carries 900 W; the same unit at 0.9 carries 1,350 W. Always check both labels — VA for apparent power, W for real continuous output. This calculator takes either: enter VA and it converts to watts using your power factor before doing anything else.[1]
The second error is structural: a single UPS is treated as resilience, even though the UPS itself becomes a single point of failure. Put one UPS in front of a critical switch, camera headend, access-control panel, or small server stack with no redundancy, and the design still fails the day the UPS fails. This article walks the watts math and the battery-derate reality the spec sheet quietly omits, then points at the redundancy architecture the watts number cannot fix on its own.
The watts math
The runtime formula is the easy part: runtime in hours equals battery amp-hours times battery voltage times efficiency times depth of discharge times the age derate, all divided by the load in watts.[2] The inputs are where projects go sideways. Amp-hours are rated at a slow discharge — usually the 20-hour rate. Real UPS discharge is far faster, and lead-acid batteries deliver meaningfully less than their published amp-hour rating at high discharge rates. Battery voltage is the string voltage — multiply by the count in the stack, not the nominal cell voltage. Inverter efficiency runs roughly 95 to 98 percent for standby topology, 95 to 97 percent for line-interactive, and 85 to 93 percent for online double-conversion — the double-conversion penalty is the price of a zero-transfer switchover.
Depth of discharge is the variable that separates chemistries and the one vendor calculators most often hide. For sealed lead-acid and VRLA, repeated deep discharge shortens life quickly, so a conservative design limits routine usable capacity to about 50 percent.[2] LiFePO4 tolerates deeper routine discharge — commonly 80 to 90 percent depending on the battery management system. Vendor-locked tools hard-code this to their own chemistry; this calculator exposes it via the chemistry input and applies the matching usable fraction, because it is the single most consequential term in the formula.
Why published runtime curves miss real installations
Most UPS datasheets quote runtime at clean reference points — typically full load and a couple of partial-load points — and assume a fresh battery, controlled temperature, and a tidy load profile. Real installs rarely sit at those references; most run between 30 and 70 percent of nameplate. The temptation is to interpolate linearly between the published numbers, and that interpolation is exactly where designs drift, because UPS efficiency is nonlinear. Efficiency droops at low load, so the energy implied by a clean low-load reference point is more than the unit will actually deliver. A straight line between two datasheet dots over-states runtime in the middle of the range where most loads live.
This is the core of what the calculator corrects. Instead of a straight line, it interpolates effective efficiency along a risertools consensus curve — a first-party MODEL, not a standard or legal authority, built by normalizing and averaging three published mid-market vendor runtime curves.[5] The model is steeper at low load, mirroring the real droop, and the result is shown beside a naive linear estimate so you can see the gap the spec sheet papers over. Treat the consensus number as a defensible planning figure, and treat the linear number as the optimistic ceiling it is — not a target.
Battery aging is real
Surveillance and IT loads cycle UPS batteries hundreds of times a year, and chemistry determines how long that cycling stays inside the spec curve. Sealed lead-acid and VRLA — the SMB default — lose capacity steadily under 24/7 float service: the planning model this calculator applies credits SLA with full capacity at year zero, about 80 percent by year 3, and about 60 percent by year 5, and high ambient temperature accelerates that loss sharply. LiFePO4 holds capacity far better — roughly 95 percent at year 3 and 90 percent at year 5 — and supports deeper usable discharge, with service life commonly past a decade. The calculator applies that per-chemistry age derate so a three-year-old SLA battery does not get credited with fresh-battery runtime; the derate table is rendered with the result so the year-by-year figures are visible beside the number.
Sodium-ion belongs on the watchlist, not in the formula. It is moving from grid-scale and pilot storage into commercial stationary-storage applications, including data-center-adjacent deployments, and battery-agnostic modular UPS platforms in the Pro class can already be configured with it alongside lithium and lead-acid. But that momentum is at the infrastructure scale; it does not make sodium-ion the default chemistry for mainstream SMB UPS products today. Treat it as a Pro-class or specialty option — substation cabinets, wide-temperature environments, maximum cycle-life cases — on platforms that explicitly support it, and watch the 2027–2028 horizon for broader SMB availability and price parity. Do not size a UPS around a generic sodium-ion cycle-life assumption unless the selected system datasheet supports it.[3]
Redundancy planning — the second error pays off
A UPS protects against utility failure, but one UPS is still a single point of failure. Borrowing data-center language: a single UPS covers utility loss but drops the load when the UPS fails; N+1 keeps enough capacity after one unit or power module fails; 2N runs two independent paths, either of which carries the full critical load. For surveillance systems, IDF closets, and small-business IT, N+1 is often the practical target when downtime has a real cost. Two orthogonal features matter across every architecture: hot-swap battery modules cut mean-time-to-repair because a degrading battery is replaced without dropping the load, and critical-load segregation puts the must-not-die circuits on a separate outlet group with deeper retention so UPS resources stretch where they count.[4]
For larger facilities the Uptime Institute tier classification gives a useful analogy, not a verdict: a non-redundant facility is commonly cited near 99.671 percent availability, a concurrently-maintainable N+1 facility near 99.982 percent, and a fault-tolerant 2N facility near 99.995 percent. The tier system applies to whole data-center infrastructure, not automatically to a closet UPS — but the gap between one UPS and N+1 is the same gap that separates the lowest tier from the middle one, and that is the gap a single-UPS design quietly ignores.
The most common SMB resilience win is UPS-backed Power-over-Ethernet: cameras, door readers, and VoIP phones survive the outage that motivated the install. A 1,000 W UPS feeding a managed PoE switch plus a handful of cameras is a light load — well inside the watts budget — but the spec still has to answer what happens when this single UPS dies. The watts number sizes the energy; the architecture sizes the resilience. Where the load mix allows it, put the must-not-die circuits on their own outlet group so the camera headend or the access-control panel keeps running after the general load has been shed.
Validate, do not trust the calc alone
The calculator gives a defensible starting number, not the final number. Four habits separate integrators who get called back from those who do not. Commission with a real load-and-runtime test: after installation, run a controlled discharge under representative load and measure actual runtime against the requirement, then correct the load profile, battery configuration, UPS size, or redundancy before closeout — not after the next outage exposes the gap. Spec capacity above the engineering minimum to absorb the year-3 derate up front; the cost delta between one UPS size and the next is rarely the line item that kills a bid. Plan a 12-month battery health review — impedance, self-test, runtime re-test, ambient check — and replace on measurement, not at failure. And document the assumptions: load profile, power-factor estimate, target runtime, redundancy architecture, and chemistry choice, so the sizing rationale is recoverable three years from now. This calculator estimates energy availability; it does not certify that a UPS can carry the load startup surge, crest factor, or power-quality requirement. The calc is the starting point; the field is where the number earns its accuracy.