The new National Planning Policy Framework, and what happened to the tilted balance.

The National Planning Policy Framework was rewritten on 17 August 2026. Not amended. Rewritten.

It is the first full rewrite since the framework was created in 2012, and it replaces the December 2024 version outright. The document that arrived is organised differently and numbered differently, and in several places it says something different. This publication has 21 articles explaining how the old one worked. This is an account of what the thing we explained now does instead.

Start with the part that has no grace period.

What applies, and from when.

The national decision-making policies in the new framework are a material consideration from the day of publication. Not from a commencement date and not after a transitional window. From 17 August 2026 (NPPF, August 2026, Annex A, paragraph 1).

That matters to anyone with an application in the system. A scheme submitted in March and still undetermined in September is now assessed against policies that did not exist when it was submitted.

There is a second rule that will do more damage. Development plan policies which are materially inconsistent with the new national decision-making policies should be given very limited weight. The only exception is where they were examined and adopted against this framework, which no adopted plan in England has been (Annex A, paragraph 2). Policies that are not inconsistent keep their weight, and the framework says so, so this is not a general invalidation of old plans. But every council with a plan adopted before December 2024 now has an argument to have about which of its policies survive.

Plan-making is treated separately. Local plans, minerals and waste plans and supplementary plans in preparation must be produced in accordance with the new framework (Annex A, paragraph 4). Neighbourhood plans get a cut-off date. One submitted to the local planning authority under Regulation 15 on or before 17 August continues under the December 2024 framework, and anything not yet submitted must comply with the new one. A qualifying body that has already submitted may withdraw its plan in order to update it (Annex A, paragraph 6).

Paragraph 11(d) does not exist any more.

Since 2018, the argument in almost every contested housing appeal in England has run through one paragraph. If a council could not demonstrate a five year supply of deliverable housing land, paragraph 11(d) engaged. The policies that would ordinarily protect a site carried reduced weight, and permission had to be granted unless the harm significantly and demonstrably outweighed the benefits. Planners called it the tilted balance. It is the mechanism that allowed 54 homes at the Wern in Lechlade in June, over an inspector's own findings of harm.

It is not in the new framework. It has not been renumbered or relocated. The mechanism itself has been replaced.

The presumption is now permanent, and it runs through the settlement boundary.

Policy S3 sets a presumption in favour of sustainable development which applies to every decision, not only where a council has failed a test. It then sends the decision one of two ways.

Inside a settlement, policy S4 applies. Development should be approved unless the benefits of doing so would be substantially outweighed by any adverse effects.

Outside a settlement, policy S5 applies, and there the presumption runs to a list rather than to development in general. Ten categories of development should be approved outside settlements, subject to the same test. Most are familiar to anyone who has read a rural local plan: agriculture and forestry, mineral extraction, infrastructure, rural business where a rural location is necessary, the reuse or replacement of an existing building, previously developed land, limited infilling within groups of houses, exception sites.

Two are not familiar, and both are new.

The first produced most of the headlines. Residential and mixed-use development within reasonable walking distance of a well-connected station should be approved outside a settlement, provided it is physically well related to the station or its settlement, is of a scale the infrastructure can take, and would not prejudice long-term comprehensive development in the same location (S5(1)(h)). The definition of a well-connected station is narrow. It must sit within one of the top 80 Travel to Work Areas in England by gross value added, and be served throughout the daytime by at least four trains or trams an hour overall, or two an hour in one direction (Annex B). Policy L3 then sets a minimum density of 35 dwellings per hectare on those sites, rising to 45 where services run at twice the minimum frequency.

For most of Gloucestershire that provision does no work at all. Lechlade has not had a station since 1962.

The second is the one that matters here.

Failing the five year supply is now a gate, not a tilt.

Policy S5(1)(j) says development which would address an evidenced unmet need should be approved outside a settlement. The policy then states in brackets what an evidenced unmet need includes: housing, where the local planning authority cannot demonstrate a five year supply of deliverable sites, or where it scored below 75% in the most recent Housing Delivery Test.

The difference from paragraph 11(d) is not cosmetic, and it cuts both ways.

Under the old framework, failing the supply test changed the weight of everything else. Protective policies were downgraded across the board, and the decision-maker then balanced harm against benefit with a thumb on the scale.

Under S5(1)(j), failing the supply test does something narrower. It downgrades nothing. It puts a housing scheme into the list of things that may be approved outside a settlement at all. Once the scheme is in the list, the test is the same test that applies inside one: approve unless the benefits would be substantially outweighed by adverse effects. Two further conditions attach. The development must be physically well related to an existing settlement, and it must be of a scale that existing or proposed infrastructure can accommodate.

So a site that sits detached from the town has an argument it did not have before, because being well related to a settlement is now a condition of qualifying rather than a factor in the balance. A site on the edge of the built-up area has lost an argument, because the question of how much weight its protective policies carry has been replaced by a question of which category it falls into.

The question is now where the settlement boundary runs.

This is the change that will consume the most officer time, and it has had the least attention.

The framework defines a settlement in its glossary. It covers cities, towns, villages and other predominantly built-up areas. It picks up areas defined as a settlement in the development plan, whether by a drawn boundary or by criteria where no boundary has been drawn. It excludes hamlets and scattered groups of houses unless the plan specifically defines them as a settlement, and it excludes villages washed over by Green Belt (Annex B).

It also includes something that will surprise people. Land which is allocated or has permission for development, and which will form part of the built-up area once that development is complete, is inside the settlement.

Read that twice. A field with planning permission on it is part of the settlement before anything is built. The boundary moves when consent is granted, not when the houses go up.

Policy S2 now requires plans to identify settlements and their boundaries (S2(1)(a)). A great many adopted plans do not. Others use development boundaries drawn for a different purpose in a different decade. Until those are redrawn, the argument moves to the criteria, and it will be argued site by site.

The standard method did not change.

Worth stating plainly, because a good deal of professional commentary assumed it would.

Annex D sets the baseline at 0.8% of existing housing stock and adjusts it for affordability, raising the baseline by 0.95% for each 1% that the five year average affordability ratio sits above five. That is the same formula the December 2024 framework introduced, and it produces the same numbers. Cotswold District's local housing need remains 1,054 homes a year.

The buffers survive and move to Annex D. 5% as standard, 20% where delivery has been significantly below the requirement. A third has been added for decision-making only: 20% where a council's adopted housing requirement was examined against a framework published before December 2024 and averages 80% or less of its current standard method figure (Annex D, paragraph 9). Cotswold District, which has been applying a 20% buffer to itself voluntarily, now has a policy home for it.

The Housing Delivery Test thresholds hold. Below 95%, an action plan. Below 85%, a 20% buffer. Below 75%, an evidenced unmet need for housing is deemed to exist for the purposes of S5(1)(j) (Annex D, paragraph 12).

Which leaves untouched the problem we wrote about four days ago. Those consequences are precise, and the last time the government published a Housing Delivery Test result was December 2024.

The 17.5% did not survive.

The December 2025 draft proposed standardising the developer profit margin at 17.5% of gross development value, setting a greenfield benchmark land value at ten times existing use value, and restricting viability assessments to unforeseen circumstances. We described those proposals as the first serious attempt to compress the numbers.

None of the figures made it into the framework. Policy DM5 carries no standardised profit margin and no benchmark land value multiple. The government has deferred standardised inputs to new planning practice guidance, to be published as soon as possible.

The direction survived. Schemes that accord with up-to-date plan policy are assumed viable. Assessments must still be published. Each one has to refer back to the assessment that informed the plan and evidence every input and assumption it uses. Neither the price paid for land nor the price set in an option agreement justifies falling short of policy (DM5(3)). The draft's restriction to unforeseen circumstances became four defined circumstances at DM5(2), which is looser than the draft and tighter than what came before it.

One rule moved the other way, and we should say so because we reported the old position. The Golden Rules on land released from the Green Belt still require up to 50% affordable housing, and substantial weight is still given to complying with them. They are no longer absolute. Policy GB8(3) admits a viability assessment in three circumstances, and the 50% requirement at policy HO5 is now qualified by the words unless this would make the development unviable. When we described the Golden Rules as the highest requirement in national policy with no viability opt-out, that was accurate on the day. It is not accurate now.

What a parish or town council should do this week.

Four things, and none of them involves reading all 130 pages.

Find out whether your parish is identified as a settlement in the adopted local plan, and where its boundary runs. That question now decides which policy applies to every application in your area, and it is the one your district may not be able to answer cleanly.

If you have a neighbourhood plan in preparation, check whether it was submitted under Regulation 15 on or before 17 August 2026. If it was, it continues under the old framework. If it was not, it must comply with the new one and needs reading against policies S6 and PM17 before it goes any further.

If you have a made neighbourhood plan, check its age. Policy S6 protects a plan that became part of the development plan five years or less before the decision is made, and only where it contains allocations meeting its identified housing requirement. Both conditions have to hold, and the second one catches plans that were written to resist allocation rather than to make it.

Stop citing paragraph 11(d) in objections. It does not exist. An objection that cites it tells a case officer that its author has not read the framework, which is an expensive way to open.

What this is, underneath.

The framework runs to 130 pages and 131 numbered policies, and most of it will take a year of appeal decisions to settle into a settled meaning.

The shape is already legible, though. The old framework turned a council's failure into a general downgrading of its own policies, which is why the argument was always about weight. The new one turns that failure into a question of category, decided against a line on a map that a great many local plans have never drawn.

We will come back to what all of this does to a speculative application on the edge of a Cotswold town, and to the 150 homes at Lechlade that have been waiting since April for a decision.