Start from the real load, not the nameplate
The most common sizing mistake is adding up the nameplate ratings printed on each device and buying a UPS to match. Nameplate figures are worst-case maximums; the actual draw of a server, switch or storage array in normal operation is usually a fraction of that. Size to the nameplate and you pay for — and cool — far more UPS than you will ever use.
The reliable approach is to measure. A clamp meter or the PDU’s own metering will tell you the true steady-state load in watts. Where a room is still on the drawing board, work from the manufacturers’ typical (not maximum) power figures and validate once the equipment is live. In a Kenyan enterprise room the grid is only part of the picture: utility voltage can sag and swell, so the UPS is doing more than carrying the load — it is conditioning the power reaching sensitive equipment.
Watts, VA and power factor
UPS units are rated in both VA (apparent power) and watts (real power). Modern IT equipment with power-factor-corrected supplies runs close to unity, so the watt figure is what matters most — but never load a UPS to its VA rating and assume the watts will follow. Always check both numbers against your measured real-power demand and keep the unit inside its rated envelope on both.
Leave headroom. Loading a UPS to 90% leaves nothing for growth, for the inrush of a device restarting, or for the day someone adds a server without telling you. A loading target in the region of 60–80% of rated capacity is a sensible working band — enough margin to be safe, not so much that the unit is oversized and inefficient.
Runtime is a design decision
Battery runtime is not a fixed property of the UPS — it is a function of how much battery you specify. The real question is: what does the UPS need to bridge? If a standby generator starts and takes the load within, say, half a minute, the UPS only has to cover that transfer plus a safety margin. If there is no generator, the UPS must carry the room long enough for an orderly shutdown of every system that cannot simply be dropped.
In Kenya, where scheduled and unscheduled outages both occur, most enterprise rooms pair a UPS with a generator: the UPS delivers instant, clean ride-through and the generator provides the endurance. Sizing the two together — UPS for the transient, generator for the duration — is almost always more economical and more reliable than trying to buy hours of battery.
Redundancy and the path to scale
For loads that genuinely cannot go dark, single-unit designs are a single point of failure. Parallel and modular topologies let you add an extra unit (N+1) so the room stays supported even with one module offline for a fault or for maintenance. Modular systems also let capacity grow in steps rather than forcing a full replacement when the load outgrows the original unit.
Whether that redundancy is justified is a business decision as much as a technical one — it depends on what an outage of this specific room costs. That is exactly the conversation a site survey is for.
A short sizing checklist
Measure the real steady-state load in watts. Check it against both the watt and VA ratings of any candidate UPS. Decide the runtime the UPS must actually bridge — to a generator, or to a safe shutdown. Add headroom for growth and inrush. Decide whether the load warrants N+1 redundancy. Then confirm the design against the site’s power quality and physical constraints before you buy.
None of these steps needs guesswork. They need a survey — and that is where a specification stops being a nameplate estimate and becomes an engineered answer.