How the reading works

It compares the quotation you upload against the items a job of that type usually itemises, and reports what is not written there.

It is not cost consultancy and it is not contract review. It has no view on your contract, your price, or the firm that wrote it.

Not listed doesn't mean not included. Some firms roll waste removal into the labour line. This is a question to ask, not a gap to assume.

Firms differ enormously in how they itemise. Some fold waste removal into the labour line. Some rely on a local convention about what a job of that type obviously includes. Some write a single line because that is how they have always priced work, and it tells you nothing about how carefully they have thought about the job.

So the output is a list of things worth asking about, not a list of gaps you may assume. This service does not use wording such as “under-quoted” or “left out deliberately”, and it infers nobody’s intent.

The service makes no judgement about whether any price is reasonable, high or low. It does not compare quotations between firms, and it recommends no contractor.

It does not look at, process or display any monetary amount. The firm’s name, its contact details and every figure are redacted before storage, and the original image is not retained. There is nothing in the system for a price comparison to be made from.

The model reads. An image model extracts the text of the items written on the redacted photograph and classifies each one against the standard list for the category. It also flags wording that leaves the scope open. That is the whole of its job.

The browser compares. The list of items a job of that type usually itemises is fixed data held on this site. Your extracted items are subtracted from it in your browser. What remains is what is not listed. It is a set difference — arithmetic on two lists, with no opinion in it.

No model is asked whether your quotation is complete, whether anything is missing, or whether anything is wrong. Those questions are never put, so they are never answered.

Site conditions, local practice and the firm’s own quoting habits are all unknown to it. Specifically:

• what is behind the walls, how level the floors are, and what the last job left behind;
• what a job of this type obviously includes where you live;
• how this particular firm writes its quotations;
• anything you and the firm have already agreed in conversation, by email or on a site visit;
• whether an item that is not written down has been priced into another line — which is the most common explanation of all.

Every reading lists these back to you, because the limits of a tool are part of its output.

Extraction from a photograph is imperfect. Handwriting, poor light, shadow, folds, unusual layouts and continuation pages all cause misreadings, and a misread item will show as not listed when it is there on the page.

Check the extracted items against your own copy before you rely on them. If something is wrong, tell us at support@irobuilding.shop — misreadings are how the standard lists get better.

The standard list for a category is a general list for a general job. If your job is specified unusually, the list will fit it badly and the reading will be worth less.

Images and text you upload are not used to train any model. Where a third-party model provider performs the extraction, the arrangement contractually prohibits training on the content we send, and we send only the redacted image. The detail is in the privacy notice.

There is no automated decision here with a legal or similarly significant effect on anybody. The service produces a list of questions for you to put to a firm. Every decision that follows is yours.

The output may not be used in a complaint, as leverage in a negotiation, in a public review of a contractor, or as evidence in any dispute, arbitral process or litigation, and it may not be relied on for medical, legal, financial, safety, employment, credit, housing or judicial decisions. The full list is in the acceptable use policy.