Maximum Demand & Diversity: What BS 7671 Actually Says

Add up the rating of every circuit in a building and size the supply to match, and you will specify an incomer two or three times bigger than you need — because not everything runs flat out at the same moment. Diversity is how you size for the realistic peak instead. What almost nobody mentions is that BS 7671 does not tell you how. It requires you to determine maximum demand, permits you to allow for diversity, and then stops.

Try it: the diversity explorer

Build a load schedule, then switch the premises type. The same loads give a different maximum demand in each column, and the explorer shows which rule fired on each category and what it took off. That spread is the point: these are conventions for a type of building, not physics.

Connected load is not maximum demand

Connected load is the arithmetic total of everything wired in. Maximum demand is what you expect to be drawn at the same instant. Diversity is the gap. In the figure above a small domestic board has 61.7 A connected; the lighting allowance takes 1.8 A off and the cooker rule takes 23.4 A off, leaving a maximum demand of 36.5 A that fits comfortably on a 40 A supply.

Note what does not move: the thermostatically controlled immersion heater. It is already cycling on its own thermostat, so there is nothing further to assume about it, and the guidance allows it no diversity at all. That pattern — long-duration thermostatic loads keeping their full value — is the single most important thing to carry into modern work, where heat pumps and EV chargers behave the same way.

What Regulation 311.1 actually says

Here it is in full. This is the entirety of BS 7671's treatment of maximum demand:

Two words carry the weight. Determining maximum demand is "shall" — mandatory. Applying diversity is "may" — permitted, optional, and entirely unspecified. There is no Regulation 311.2; Chapter 31 moves straight on to conductor arrangement. If your demand assessment turns out to be wrong, the failure is against 311.1's "shall", not against any table.

And the standard is no more forthcoming elsewhere. In 642 pages the word "diversity" appears fifteen times, and not once with a number attached. There is no diversity table. There is no "current demand to be assumed" table. There is no cooker allowance. "Maximum demand" is not even a defined term in Part 2 — though "diversity" is, circularly, as a "means of determining maximum demand … taking into account usage patterns".

Most tellingly of all: BS 7671 never once tells you where to look. There is no cross-reference to the On-Site Guide or to any Guidance Note anywhere in the book. Appendix 1 lists British Standards; Appendix 2 lists statutory instruments and HSE publications. The industry's universal reliance on the On-Site Guide's table is a convention supported by nothing in the standard's own text.

So where do the numbers come from?

From IET On-Site Guide, Appendix A — "Maximum demand and diversity" — and its Table A2, "Allowances for diversity". The equivalent material is in Guidance Note 1. Both are IET publications sold alongside the standard. Neither is the standard.

That would be a pedantic distinction if the guide were confident about its own figures. It is not. Appendix A opens by saying the values are "intended only for guidance, because it is impossible to specify the appropriate allowances for diversity for every type of installation and such allowances call for special knowledge and experience", and then goes considerably further:

Put the two documents together and the position is clear. The standard requires an assessment and gives no method. The guide gives a method, disclaims it, and declines to cover a large fraction of the buildings people actually design. Diversity is a judgement you own. You can defend it with reasoning, load profiles and metered data. What you cannot do is point at a regulation and say the number came from there.

How the allowances actually work

The most common mistake in software — including, until this week, our own calculator — is to treat diversity as one percentage per category. It is not. The guide's allowances have three different shapes, and only one of them is a plain percentage:

• A straight percentage of the group. Lighting works this way, and the percentage differs by premises type — roughly two thirds in a house, nine tenths in a shop or office, three quarters in a small hotel. Offices are lit consistently all day; houses are not.
• Ranked: the largest in full, then a decreasing share of the rest. Most commercial rows work this way — the biggest appliance or motor counts fully, the second at a reduced fraction, the remainder at a lower one again. It is a bet that the largest single item is the one running.
• Banded: everything up to a threshold, then a fraction above it. Domestic heating and other unlisted final circuits take the first 10 A in full and half of anything beyond, so a small load gets no reduction at all and a large one gets a substantial one.
• A fixed base plus a share of the excess. The household cooker is the famous one: the first 10 A, plus 30 % of the connected rating above that, plus a further 5 A if there is a socket-outlet in the cooker control unit. A 43.5 A cooker comes down to 20 A.
• Nothing at all. Thermostatically controlled water heating and floor warming get no diversity in any column. They are already self-regulating; there is no second assumption to make.

None of those can be expressed as a single multiplier, which is why a calculator that uses one will disagree with a hand calculation. Our maximum demand calculator now applies the staged rules and shows the working for each category, so you can see which rule fired rather than being handed a number.

Same schedule, three answers

This is the part worth internalising. The guide gives three columns — individual households including individual dwellings of a block; small shops, stores, offices and business premises; and small hotels, boarding houses and guest houses — and the same load schedule can produce wildly different maximum demands depending on which one you use.

The 61.7 A domestic board above comes out at 36.5 A on the household column and just over 61 A on either commercial column: the difference between a 40 A incomer and a 63 A one. The reason is the cooker. The household rule bands it down to 20 A; the commercial rules rank it, and with a single cooker "100 % of the largest appliance" means no reduction whatsoever.

Choosing the wrong column costs more than choosing the wrong method — and remember the guide offers no column at all for a block of flats, a large hotel, a factory or a large commercial building. If that is your job, the table does not cover you and the assessment is entirely yours.

A worked example

Diversity sizes the supply, never the board

This is where Amendment 4 quietly tightened things, and it is the half of the subject that almost never gets written about. Regulation 536.4.202 gives three routes to protecting a low-voltage switchgear and controlgear assembly — which the regulation's own note confirms includes a distribution board and a consumer unit — against overload. One of them is:

So the assessment you use to size the incomer is not the assessment that sizes the board. If you take that route, the board carries everything bolted to it. The alternatives are that the upstream protective device is rated at or below the assembly's rating, or that real load curtailment limits the demand — and curtailment means an engineered control, not an assumption.

Note the circularity in the one permissive case: Regulation 536.4.3.2 is the only normative sentence in BS 7671 that uses the phrase "diversity factors", and it permits them "according to Regulation 311.1" — which supplies none. Regulation 551.7.2.2 repeats the "without diversity" rule for an assembly with a generator in parallel, and requires a warning notice stating the maximum permitted connected load.

The switchboard's own "diversity factor" is a different thing

Assemblies to BS EN 61439 carry a rated diversity factor, and it trips people up because it wears the same word. The standard defines it as:

"the per unit value of the rated current, assigned by the assembly manufacturer, to which outgoing circuits of an assembly can be continuously and simultaneously loaded taking into account the mutual thermal influences"

Read that carefully. It is a thermal limit on the enclosure — how much of each circuit's rating can be drawn at the same time before the assembly overheats — and it is declared by the manufacturer, not chosen by the designer. It says nothing at all about whether the loads will in practice run together. Expressed as a formula it is simply the operating current over the circuit's rated current, and the default values by number of circuits are tabulated in BS EN 61439-2. Read them there; we do not reproduce figures we have not verified against the source.

The two quantities are orthogonal, and the failure mode is treating the assembly's rated diversity factor as diversity you have "already claimed", or vice versa. They answer different questions and both have to be satisfied.

EV charging, heat pumps, and why the old habits now bite

The diversity table was written for a world of intermittent loads. Two modern additions break its assumptions, and the standard has responded to only one of them.

• EV charge points — Reg 722.311.201. Section 722 does not offer a diversity allowance. It permits load curtailment — "load reduction including disconnection, either automatically or manually" — to be taken into account when determining maximum demand. Combined with 536.4.202's prohibition on using diversity as a means of load curtailment, the conclusion is unavoidable: you may install a load-management system and count it, but you may not simply assume the chargers will not coincide.
• Heat pumps — nothing at all. BS 7671:2018+A4:2026 does not mention heat pumps anywhere: not a section, not a regulation, not an index entry, in a standard amended in 2026. The On-Site Guide's table predates them. A heat pump is a long-duration thermostatically controlled load, so the closest analogous row is the one allowing no diversity — treat it as near-continuous unless you have real data.
• Everything else that runs all day. Server rooms, commercial refrigeration, electric boilers. The common thread is duration: diversity is a bet about coincidence, and loads that run for hours coincide.

This is why so many boards that were "adequate" now need upgrading. Adding a 32 A charge point to an installation is not adding 32 A of diversified load — it is adding 32 A.

On three phases, balance matters as much as total

A three-phase maximum demand is not one number. The incomer and the tails are sized on the governing phase — the busiest one — so an installation can be well within its supply on average and over it on one line. The explorer above reports the per-phase figures and the imbalance percentage alongside the total.

Two consequences. First, moving one large single-phase load between lines can reduce the maximum demand without changing the connected load at all — the cheapest capacity you will ever find. Second, imbalance drives the neutral current, which is its own design question once harmonics are involved: see the derating factors article on triplen harmonics, where above a third-harmonic threshold you size on the neutral rather than the line.

ADMD is a different quantity

After Diversity Maximum Demand is what a distribution network operator uses across a group of dwellings, quoted as kW or kVA per property. It diversifies across many households whose peaks fall at different times, so it falls as the number of properties rises — the opposite direction from the installation diversity above, which is fixed by category.

They are not interchangeable. Sizing one property's tails on an ADMD figure will undersize them; sizing a network on one property's installation diversity will oversize it substantially. And note that the On-Site Guide explicitly declines to give guidance for blocks of dwellings — exactly the case where people reach for ADMD instead.

What it costs to get it wrong

Maximum demand sizes the incomer, the main tails, the distribution transformer or standby generator, and the DNO connection. Overestimate it and the client pays for capacity and copper they will never use. Underestimate it and the main protection operates, or the connection is too small to extend later — and a DNO upgrade is one of the slowest and most expensive changes to make after the fact.

The design current that comes out of this assessment is the I_b that every downstream check consumes: the five checks of a cable calculation, voltage drop, and the rating factors applied to current-carrying capacity. Get demand wrong and every one of them is wrong with it. If you are starting from loads in kilowatts, the kW to amps calculator will get you to the currents this assessment starts from — it deliberately does not apply diversity, because that is the judgement described on this page.

Glossary

Frequently Asked Questions

Sources and verification. BS 7671:2018+A4:2026 Regulations 311.1 (maximum demand and diversity), 132.1 and 132.3 (design and nature of demand), 132.16 (additions and alterations), 313.1(e) (suitability of the supply), 433.1.204 and Appendix 15 (standard circuit arrangements, informative), 523.9, 536.4.3.2 and 536.4.202 (assemblies), 551.7.2.2 (parallel generation) and 722.311.201 (EV load curtailment); Part 2 definitions of diversity, design current, load curtailment and load shedding. Diversity allowances are IET On-Site Guide Appendix A, Tables A1 and A2 — guidance, not the standard, and quoted here in summary rather than reproduced. The rated diversity factor is BS EN 61439-1 §5.4, with its default values in BS EN 61439-2 Table 101, which we have not reproduced. Verify every allowance against your own copies before issue; the numbers on this page are indicative and, as the guide itself notes, may not align with modern loads.