The Building Safety Regulator’s Gateway 2 numbers are climbing, but the regulator is clear that the rise comes from better applications, not a lower bar. Read the other way, roughly one in four higher-risk building submissions is still rejected, and even the ones that pass are paying for it earlier than ever.
The numbers, read the other way.
In its data to 28 June 2026, the BSR put Gateway 2 approvals at 77% overall (GOV.UK), and the trajectory is real. A dedicated fast-track unit now handles most live new-build cases, the regulator cleared more than 700 decisions in the final quarter of 2025 against roughly 200 in the first, and since January 2026 it has operated as a standalone body (GOV.UK; RIBA). That is genuinely encouraging, and the improvement is attributed to higher-quality applications rather than a relaxed standard.
Turn the same figure around, though, and the harder truth tells a different story. About a quarter of higher-risk submissions are still failing at the gate, and the BSR’s own goal of turning around non-complex applications within 18 weeks is not due to be met until March 2027. Faster is not yet fast, and passing is not yet routine.
Gateway 2 approvals overall in the BSR’s data to 28 June 2026, which means roughly one in four higher-risk building submissions is still rejected.
Why a quarter still fail.
Rejections come from two overlapping causes. Sometimes the information is incomplete, inconsistent, or not structured enough for the regulator to assess against the requirements. Sometimes the design itself is not resolved or not compliant enough at the point of submission. More often than not it is both at once, and the two feed each other. A design that is not properly nailed down cannot be evidenced cleanly, and a gap in the information is frequently a gap in the thinking.
Interpretation makes it harder still. Views differ on how prescriptive guidance such as BS 9991 should be, and on the limits of Regulation 7 and “zero combustibles”. A point accepted by one case officer or fire engineer can be read differently by another, and there is a persistent fear that a position settled at Gateway 2 could be reopened at Gateway 3. Either way, a quarter of submissions still are not getting it right first time.
Passing is now front-loaded.
The firms that clear the gate pay for it earlier. To pass, the design is locked in much sooner, with more time and cost committed up front, and decisions are made before their downstream consequences are fully understood. Get that wrong, and you are committed to a design that is not resolved enough, and unwinding it later, mid-construction or through a change-control application, is the expensive way to find out.
Whether you’re trying not to be the one in four, or carrying the front-loaded cost of being the three in four, the same thing separates the two: submissions whose completeness, consistency and reasoning are repeatable, not heroic.
Four kinds of knowledge a submission has to control.
Underneath the two failure modes is a simpler idea about what a submission has to get right. It has to control three kinds of knowledge, and each one needs handling differently.
| Knowledge | What it is | What it needs |
|---|---|---|
| Known knowns | The requirements you know you must meet. | Full traceability: every decision linked to its rationale, its author and the requirement it serves. |
| Known unknowns | The items you know are still open. | Flagging and closing before they reach the regulator, not after. |
| Unknown Knowns | Requirements which are known but their significance is not appreciated (a past lesson, a buried standard, one person’s judgment). | A searchable, proactive record so your own experience reaches the submission without bias or assumptions. |
| Unknown Unknowns | The clash, omission or assumption nobody thought to check. | Only caught if something has seen it fail on a comparable project before. |
Rumsfeld’s Matrix: Understanding Rework in Construction / Characterizing unknown unknowns
Seen this way, an incomplete-information rejection is usually a known unknown left open, and the expensive, late-stage failures are the blind spots that nothing surfaced in time. A submission that controls all four is what “right first time” actually means.
Faster to review, not just easier to pass.
There is a second prize here that is easy to miss. The average time to approval is 22 weeks against about 6 weeks for a rejection for a new higher-risk building at Gateway 2. The review process is not spent doubting the design. Much of it is spent reconciling the submission, chasing missing documents, resolving drawings that do not match the specification, and requesting the evidence a claim needs. Every such request stops the clock, and in practice applications have run 25 to 40 weeks (Walker Morris).
The regulator has been explicit about what helps them. An application that is coherent and easy to follow, where key information can be located and verified without unnecessary cross-referencing. That is a documentation problem as much as a design one. When every requirement is mapped to the evidence that satisfies it, when the model, drawings and specification are reconciled before they are sent, and the reasoning is traceable, the reviewer spends their time confirming compliance rather than hunting for it. A structured, cross-referenced, auditable submission is not only more likely to pass. It is faster to sign off, on both sides of the table.
The wider picture.
None of this sits in isolation. Attention is already turning to Gateway 3, where thousands of completed units have sat unoccupied across delayed schemes, though the BSR notes that many are legacy cases and that no reformed-Gateway-2 new-build has yet reached Gateway 3, so the real test is still ahead (Building). The development model has front-loaded too, with the whole supply chain engaged early, forward funding held back until Gateway 2 approval, and the Building Safety Levy arriving in October 2026. And it lands as the industry is shortest of the people it needs, with building control capacity down sharply and fewer than 30 fully competent fire engineers in England’s fire and rescue services (NFCC). Every one of these pressures rewards the same thing, firms that can control their requirements and evidence process, and stop relying on the right person being in the room.
From checking rules to assuring your judgment.
Both failure modes point at the same capability gap, and closing it opens three concrete opportunities:
- Make traceability automatic. The golden thread (GOV.UK guidance) produced as a by-product of the work rather than a manual scramble before the gate. See requirements management.
- Reconcile the package before it is sent — model, drawings and specification made consistent, so the mismatches that trigger a request for information are caught first.
- Assemble a submission the regulator can read fast— every requirement mapped to its evidence, coherent and navigable, so validation is quicker, fewer applications are rejected as incomplete, and sign-off comes sooner.
- Forecast, don’t react. Use past and live projects to predict the impact of a decision before it is locked in, while changes are still cheap.
- Flag changes against the gate. After approval, flag automatically whether a change needs a planning amendment or, under the Building Safety Act’s change-control regime, full BSR approval before work proceeds, so a late deviation can’t quietly invalidate what was agreed.
The deeper shift underneath those five is what really matters. It moves you from checking a design against a set of rules to assuring your own requirements and judgment. The reasoning behind each decision, not just its compliance status. Done this way, every submission builds a firm-specific record of resolved judgment, one that is yours alone and cannot be copied, and it replaces a binary pass or fail with graded, calibrated confidence in the outcome. A human still signs the record, but they sign faster and with confidence.
Done this way, every submission builds a firm-specific record of resolved judgment, one that is yours alone and can’t be copied, and replaces a binary pass-or-fail with graded, calibrated confidence in the outcome. It is a system that makes completeness, consistency and reasoning repeatable.
Quality at that level is not heroics. It has to be systematic, and it has to compound.
The BSR’s rising numbers are good news for the industry. But they also mean the bar is now a resolved design and the information to prove it. The firms that make “right first time” repeatable will pass faster, carry the front-loaded cost more cheaply, and pull away from those still treating each submission as a fresh scramble.
See it on your own package
If Gateway 2 and 3 sits on your risk register, talk to Tektome about assuring your own requirements and judgment, not just checking against rules. Every submission builds a firm-specific record of resolved judgment that’s yours alone and can’t be copied, with graded, calibrated confidence in place of a binary pass or fail.