Add up everything that could draw current in one house and you get a number no house ever reaches. Add up eight houses and the gap between arithmetic and reality becomes the difference between a sensible supply and an absurd one. Diversity is the engineering answer to that gap, Regulation 311.1 requires it to be assessed, and it is the assessment most often done by habit rather than method.
The method has two layers, and keeping them separate is most of the job.
Layer one: diversity within the dwelling
The connected load of a single dwelling is not its demand. The Design Guide's allowances, from the familiar On-Site Guide lineage, price each load type:
Lighting takes 66 percent of its total current demand. General heating and power takes the first 10 A in full plus 50 percent of everything above. The cooker takes the first 10 A, plus 30 percent of the remainder, plus 5 A if its control unit has a socket, and that allowance is applied once, where the loads are summed, not again on a figure that already carries it. And a short list of loads gets no diversity at all: thermostatically controlled water heaters, floor warming, and thermal storage heating, because a thermostat holding temperature is a load that shows up whenever it likes for as long as it needs.
Run a typical dwelling through those rules and the connected load collapses to an assessed maximum demand, commonly somewhere around 60 A for a standard house without electric heating.
Layer two: diversity across dwellings
Eight houses at 60 A do not make a 480 A supply, because eight households do not peak together. The Design Guide handles this with Figure 3.4: a demand factor g, read against the number of dwellings, on one of three curves chosen by how coincident the households' usage is.
The Design Guide describes them in one line: curve a is high coincidence of usage, curve b typical, mixed domestic loads, and curve c low coincidence. It recommends none of them.
At 8 dwellings curve b gives 0.22, the one value the book prints as a number; read off the chart, curves a and c sit at about 0.43 and 0.12. The factor is applied to the dwellings after the first, which is taken in full: on curve b the eight houses at 480 A of summed maxima become 60 + 0.22 × 7 × 60 = 152 A. Roughly a third of the naive total, delivered entirely by the observation that people do not cook, shower, and charge everything simultaneously across a whole street.
The decision hiding in the graph
Look at the three values again: about 0.43, 0.22 and 0.12. The spread between curve a and curve c at the same dwelling count is a factor of more than three, and it carries through to the supply: about 241 A on curve a against 110 A on curve c for the same eight houses. The dwelling count is a fact; the curve is a judgement, and it is the single most consequential judgement in the calculation.
The guidance is directional: coincident usage, similar households on similar routines, or shared electric heating pushing everyone's demand into the same cold evening, points at a. A genuinely mixed development points at b. The judgement belongs to the designer, and it belongs in the design record with its reasoning, because a checker who cannot see why curve b was chosen cannot check the thing that actually decided the answer.
Diversity also runs one way. It sizes the supply down; it never excuses a final circuit. Every circuit inside each dwelling is still designed for its own full duty. Diversity describes the crowd, not the individual.
The modern loads that break the curves
Those curves encode decades of observed domestic behaviour, and two loads are busy rewriting it. A heat pump in a cold snap behaves like the no diversity category: every unit on the street working at once, all night. An EV charger pulls a large, flat, hours long load that lands in the same evening window as everyone else's. Neither behaves like the household the curves were drawn from, which is why heat pump and EV heavy developments deserve deliberate demand assessment rather than an inherited g value, and why the curve choice question is getting harder rather than easier.
A detail from implementing this
Building this into CurrentProof, the design decision that mattered most was refusing to let the two layers touch. The within dwelling allowances and the across dwellings curve are separate stages with a visible number between them, the per dwelling assessed demand, because the classic failure in manual diversity calculations is blending the layers and diversifying twice. The result flatters the supply, nothing flags it, and the error only surfaces years later as a main switch that runs warm every winter evening.
If your spreadsheet goes from connected loads to a final amperage in one step, that is the place to look.