How grades work

We do not judge whether a defect was fixed to a reasonable standard, and we do not decide how many defects a job of a given size ought to produce. Both are guesses, and the second one is worse than a guess: a client who inspects every week logs many more defects than one who visits twice, on identical work.

So we measure three things that are entirely about the firm: what it did with the list it was given.

The share of a client's defects the firm actually resolved by the end of the project, weighted so that serious defects count for more than cosmetic ones.

A defect's weight is set by the kind of work it is:

• Painting / touch-up ×0.5
• Tiling ×1
• Glass & aluminium ×1
• Plumbing ×1.2
• Electrical ×1.2
• Wet works ×1.2
• Hacking / masonry ×1
• Carpentry ×1.5
• Waterproofing & structural ×2
• Other ×1

A firm that fixes every paint mark and leaves a leak has resolved most of the items and very little of the weight, and its figure reflects that.

A defect settled by agreement with the client counts as resolved. There is no limit on how many can be settled and no penalty for settling.

Anything put right within 4 days is prompt and scores full marks. From there the score falls away, reaching zero at 30 days. A defect never resolved counts as the full 30.

Not every defect needs doing straight away, and a client and firm often agree that something can wait until later in the works. So each defect carries a “needs attention now” switch, on by default, which only the client can change.

• On. The defect is measured against 30 days.
• Off. It is measured against the length of the job instead.

Switching it off does not remove a defect from the measurement. It gives it a longer clock, which is what was actually agreed. A deferred defect still sitting unresolved at the end uses its whole allowance, exactly like any other.

Every time the switch is flipped, it is logged with who and when, and the firm sees it. Once grades are published, the share of a firm's defects that were marked as needing attention will appear alongside the rest of its record.

Of every repair the firm claimed to have made, the share the client accepted on the first attempt. A fix that was claimed, rejected, and then abandoned counts here as a failure.

Because it is measured against what the firm attempted, we always publish it with that wording (“of attempted fixes held first time”) so it is never mistaken for a statement about the whole list.

Each finished project gets a letter from A to E. A firm's grade is three of those letters, read left to right:

1. Typical project. The middle one of the last 10. With an even number, the lower of the two middle ones.
2. Average. Every project counted, the bad ones too. This is the letter one bad job pulls down.
3. Most common. The letter most of those projects got.

A firm that is excellent every time is AAA. A firm that is usually excellent with one or two bad jobs reads ABA: its typical project is still an A, and the average says what happened.

• Fewer than 3 finished projects: no grade yet (- - -). That is an empty record, not a bad grade, and it stops one unusual job from deciding a letter.
• 3 or 4 projects: the first two letters, and a dash for the third (for example AB-).
• 5 or more: all three. If two letters are tied for most common, the third stays a dash.

• A
  0.85 and aboveAlmost everything resolved, quickly, and the fixes held
• B
  0.75 and aboveAbout what most firms manage
• C
  0.62 and aboveNoticeably below most firms
• D
  0.45 and aboveWell below
• E
  below 0.45Most defects left unresolved, or left for a long time

Everything above is measured except three numbers, which we picked:

• 4 days counts as prompt, to become the speed of the quickest quarter of firms on the platform.
• 30 days scores nothing on speed, to become the speed of the slowest tenth.
• 14 days before sign-off is too late to count, to be reviewed once enough projects have finished.

The cut-offs between letters are provisional. They are reasoned rather than fitted, because no firm has finished enough projects to fit them against. They will be set from the real spread of firms before any grade is published, and we will say on this page when that happened and how many projects it was based on.

Three people can add a defect to a project: the client, the defect checker the client appointed, and the firm itself. Only the client can send the list, accept a fix, settle a defect or sign off. A firm can take back a defect it raised itself, and the client is emailed and notified the moment it does. Nobody can delete a client's defect but the client.

A defect a firm raises about its own work is scored against the firm, never for it. Left unresolved it counts against the firm exactly like one the client found. Resolved, it is taken out of both halves of the published figures, so self-reporting cannot lift a firm's numbers. It still gets the item written down and fixed.

We show the split on every firm page: how many defects clients raised, and how many the firm raised about itself. A defect your appointed checker raises counts as one you raised, because they are checking on your behalf. Who raised a defect is recorded once, when it is added, and nobody can change it afterwards, including the firm and us.

The grade is computed from the record and nothing else. We never adjust a grade by hand and there is no number anyone here can reach in and change. If something in the record is wrong, correcting the record recomputes the grade on the next nightly run.