Inside the Environmental Report: the free screen every survey gets, and the paid search beside it

Environmental risk on this platform is two products, and confusing them is the fastest way to misprice a job or over-promise to a client. One is a free screen that Home runs against every property, drawn from named public datasets, which flows into the report you are writing. The other is the Environmental Report, a specialist third-party search you order in one click for £75 including VAT, which arrives as a PDF filed against the property and sits beside your survey. This post explains how each one works, where they meet, and where they deliberately do not.

The free screen runs whether or not you buy anything

Every property lookup on Home carries an environmental risk fetch. It covers radon; ground stability and coal mining; flood from rivers and sea as well as surface water; landfill and air quality; noise from roads and railways; and invasive plants such as Japanese knotweed. That same fetch surfaces in three places: the property page a consumer sees, the property data drawer beside a report you are drafting, and the report pipeline itself.

The screen is not a table we bought once and shipped. It is a live call to Home's own data service, which in turn draws on the public providers the code names:

There is a useful piece of history in the code here. An engineering note records that the upstream ground stability and radon calls once ran to eight or ten seconds each, and were pulled out of one consumer-facing page for that reason. That is the tell that these are external requests to live services rather than a cached national extract sitting in our own database. The energy performance figures that appear alongside them come from a different route again, read from the facts already established earlier in the report rather than fetched a second time.

For a surveyor, the practical value is before the inspection rather than after it. Seeing a surface-water flag or a coalfield postcode while you are still pricing the job changes how you plan the visit. An address can sit nowhere near a watercourse and still carry a surface-water flag, which is exactly the kind of thing the Environment Agency's long-term flood risk service exists to make visible.

Eleven labels, three buckets, no local re-grading

Each risk comes back as a small record: a severity label, a score with its unit, whether the hazard intersects the property, and the distance to it in metres. Home passes those through as given. There is no threshold of ours sitting on top, no house scoring model that quietly promotes a provider's Low to a Medium.

The severity scale has eleven labels. For display, the platform groups them into three buckets:

The eleventh label is Not detected, which sits outside the buckets because it is an absence rather than a grade. The buckets exist so an interface can colour a row, not so a verdict can be reshaped. If you need the precise label a provider returned, it is still there underneath.

What the report writers are allowed to say about it

Environmental data reaches a survey through three sections, and the rules around it are tighter than the rest of the drafting pipeline.

Section C carries the local environment description. Section H carries other matters for the client's legal adviser, which is where an environmental finding usually needs to end up if anyone is going to act on it. Section I carries risks.

The AI section writers that draft those passages are constrained to the data that came back. They are instructed not to invent flood or mining wording that the feed does not support, which matters more than it sounds. A language model asked to write about a property near a river will happily produce a plausible paragraph about flood history from nothing at all. The constraint is that a screen returning Very Low for rivers and sea produces prose that says so, not prose that hedges its way into implying a problem. This is the same principle that governs the rest of the drafting stack, described in how the Otto report pipeline works: the deterministic data outranks the model that writes about it.

The provenance statement, and the advice it gives against us

Section I closes its environmental content with a provenance statement. It names the providers behind each risk family and records the date the data was retrieved, so a report read two years later shows what was known when, rather than presenting a stale figure as current.

The same statement recommends that where a risk looks material, a full environmental search from a specialist provider should be obtained. It names the sort of report it means. That recommendation is written into the tool by design. A screen tells you whether to look harder. It does not discharge the duty to look.

On freshness, the platform draws a line between browsing and writing. The property data drawer holds a risk lookup for an hour, because a surveyor clicking around a shortlist should not fire the same external call repeatedly. The fetch that feeds a report being written is not cached at all and runs fresh every time. How often the underlying national datasets are republished is a matter for the providers, and we do not assert a figure for it.

The paid search: what the one click sets off

The Environmental Report appears as a quick-start tile on your Reports page, with a deep link if you want it bookmarked. You pick the property once, and on the wallet billing stack the price is a flat £75 including VAT. Some older organisations still pay by card, where the price is held per product in the platform rather than as a fixed figure in the code, so the number you see at checkout is the one that applies.

What happens next is worth understanding, because it is where the accuracy of the search is decided.

Home tries to tell the provider exactly which plot you mean. First it looks for the Land Registry title polygon for the property, allowing a short budget for that lookup. If a polygon comes back, the order is placed against the boundary of the actual plot. If it does not, the order falls back to a coordinate. If there is no usable coordinate either, it falls back to the address, and the platform logs a warning against that order, because an address string is the weakest of the three and we would rather know when we used it. A polygon distinguishes a garden that reaches the floodplain from a rooftop point that does not.

The order is placed once, even if the queue tries twice. The job that places the order takes a short lock, then checks whether an order number is already recorded against that report before it does anything else. If one is, it stops. The job retries a small number of times with a minute between attempts, and the outbound request carries its own timeout, so a slow provider produces a retry rather than a hung job. Idempotency here is not a nicety. Without it, a retried job is a second £75.

Delivery arrives by webhook, not by polling. When the assessment completes, the provider calls Home back. That call carries a bearer token which is verified before anything else happens, and the handler takes a lock so two notifications for the same order cannot both act. The PDF is downloaded only from an allow-listed host, then stored on private report storage under a path keyed to the report. It is not written to a public bucket with a long URL. The card in your Reports list flips to done and you are notified.

What it will not do

Honesty about limits is more useful to a surveyor than another feature claim, so here are four.

We do not promise a delivery time. The provider has its own turnaround expectations, and you will see them quoted. Home does not track, time or hold anyone to them, and there is no timer in our code that would let us. Treat any turnaround figure as the provider's statement rather than ours.

The paid PDF does not rewrite your survey. This is the distinction people most often get wrong. The free screen feeds the report: it lands in Sections C, H and I as described above. The purchased PDF is never read back into the survey. It is filed against the property and delivered to you, and what goes into your report from it is what you decide to put there. That is deliberate. A third-party document should not be able to edit a professional report carrying your name on it.

Home runs no planning-constraints check of its own. Planning appears in the third-party product's description of what it covers. There is no Home client, dataset or table behind it, so nothing on the free screen speaks to planning constraints, and you should not read the absence of a planning flag on our side as a clear result.

A screen is not a search. The free layer is a triage instrument built from national datasets at property resolution. It is good at telling you which of six risk families deserve attention on this address. It is not a specialist environmental search, it does not carry the interpretation or the recommendations a full report does, and it is not what a conveyancer means when they ask whether an environmental search has been done. That is the gap the paid report fills, and it is why the tool's own guidance in Section I points you at one rather than pretending the screen is enough.

Section I closes its environmental content with a provenance statement.

Why it is built this way

The design follows from a simple observation about liability. The parts of environmental risk that can be established from public data should be free, present on every job, and never quietly re-graded by us. The parts that require a specialist assessment should be a deliberate purchase, ordered against the most precise geometry available, delivered privately, and left for the surveyor to interpret.

That split runs through both Level 2 and Level 3 reports, and it is the same instinct behind the risk data we show consumers on an impartial property page. Show what the public record says, name where it came from, date it, and be clear about the point where a screen stops and a search begins.