DPGF breakdown by trade with ChatGPT: method and checks
DPGF by trade with ChatGPT: why an incomplete row must stay visible, the most common VAT error and the pack’s five sheets.
By Educasium
Install the ChatGPT skill: DPGF breakdown by trade →
An architect asks ChatGPT to build the DPGF (lump-sum price breakdown) for a painting trade from the CCTP and a partial takeoff. Item P01, 10 m² at €25/m² excluding tax, is complete: the calculation gives €250 excluding tax with no ambiguity. Item P02, however, has no known quantity at this stage — the substrate has not yet been confirmed. The riskiest reflex would be to let ChatGPT fill that line with a plausible estimate so the table "looks finished"; the right reflex is to leave the row visibly incomplete, with its status displayed as such.
This article details why a DPGF must make its incomplete rows visible rather than hide them, the most common VAT error when a price breakdown mixes several rates, the method for building this document with ChatGPT without ever turning a blank cell into zero, and the situations where this breakdown is not enough.
Summary
- A DPGF is not a table that must always look complete
- The most common VAT error in a price breakdown
- Building a DPGF with ChatGPT, step by step
- A table of the pack's five sheets and their role
- When this breakdown is not enough
- Training to make DPGF documents more reliable
- Frequently asked questions
A DPGF is not a table that must always look complete
The lump-sum price breakdown (DPGF) details, item by item, the price of every service in a trade, consistent with the articles of the matching CCTP. Its value comes not from looking complete, but from being reliable: a table showing a price on every row, including rows whose quantity or unit price are still unknown, gives a false sense of certainty that backfires as soon as the actual pricing diverges.
Why an incomplete row beats a guessed one
An item with no known quantity, left visibly incomplete in the summary, tells anyone reviewing the document that a figure is still needed before final pricing. The same item filled in with a plausible estimate, not flagged as such, disappears from the reviewer's radar — until the gap with the consulted contractor's actual price reveals the error, usually too late to anticipate it.
The most common VAT error in a price breakdown
A DPGF mixing several trades or several types of work often falls into the same trap as bid comparison: applying a default VAT rate without checking whether it actually matches each item.
The rate never depends on the RGE certification of the contractor whose price was used
The 10% VAT rate, applicable to improvement, transformation, fitting-out or maintenance work on a dwelling completed over two years ago, never depends on the RGE certification of the contractor whose price served as the reference for an item. Since 1 March 2025, the only administrative condition for applying that rate is a statement on the quote or invoice, as detailed on the impots.gouv.fr page on VAT rates for housing works. A DPGF applying 20% reflexively to every item, or 10% without checking the dwelling's actual age, distorts the total even though the scope of work has not changed.
| Nature of the item | Possible VAT rate | Main condition | RGE status required? |
|---|---|---|---|
| Renovation of a dwelling completed over two years ago | 10% (or 5.5% for a defined list of energy-saving works) | Dwelling's age, statement on the quote | No |
| New construction or significant extension | 20% | No age condition | Not applicable |
| Work not eligible for the reduced rate | 20% | — | Not applicable |
Building a DPGF with ChatGPT, step by step
Step 1: Gather the CCTP, dimensioned drawings or an explicit takeoff, and any prices already known. Without these documents, ChatGPT can only produce an empty template with no real link to any CCTP article.
Step 2: Create one row per priced service, explicitly linked to its CCTP article. This article-to-row correspondence must be checkable afterwards, never merely implied by the row order in the table.
Step 3: Fill in unit, quantity, takeoff method and status for every row, without completing missing data. An item with no known quantity stays explicitly incomplete in its displayed status, never filled with a plausible value just so the row looks finished.
Step 4: Confirm the applicable VAT rate item by item, based on the actual nature of the work. This rate is never inferred from a default value reused without checking, nor from the certification of a contractor whose price was used as a reference.
Step 5: Separate options from base scope into distinct columns, then recalculate totals in Excel. An option conflated with the base price artificially inflates the total and makes the DPGF harder to compare with bids received later.
Fictional example for testing the method: item P01, 10 m² at €25/m² excluding tax; item P02 has no quantity. The first total must be €250 excluding tax. The correct request to ChatGPT is to leave the second one blank and flag an incomplete item in the summary, without ever applying an unspecified VAT rate or inventing a plausible quantity.
A table of the pack's five sheets and their role
| Sheet | Role | What it must never contain |
|---|---|---|
| DPGF | Item-by-item detail, linked to CCTP articles | A row filled with an unflagged assumption |
| MÉTRÉS | Quantities and calculation method for each item | A quantity copied from an earlier project without checking |
| SYNTHÈSE | Trade-level totals, recalculated from the detail | A total rewritten by hand, disconnected from the detail's formulas |
| SOURCES | Origin and date of every price used | A price with no reference or date |
| À_CONFIRMER | List of incomplete or pending items | A row silently migrated into the DPGF without confirmation |
When this breakdown is not enough
A well-built DPGF does not resolve every situation encountered when pricing a trade.
A project whose exact nature of works remains uncertain — a heavy renovation that could tip into reconstruction for tax purposes, for example — requires a preliminary qualification of the applicable VAT regime, possibly with the help of an accounting professional, before any detailed breakdown. A significant gap between the DPGF and bids received from consulted contractors should prompt a review of the quantity and price assumptions used, rather than a mechanical adjustment of the total to match the bids. A DPGF prepared with AI assistance, ChatGPT like any other tool, remains in every case a draft: it must be reviewed and approved by the quantity surveyor or the architect responsible for pricing before it is circulated to contractors.
Training to make DPGF documents more reliable
A well-designed pack is not enough on its own: it also needs rebuilding for every trade, with prices and quantities that change from one project to the next. For an architect practising as a self-employed professional (NAF code 7111Z in France), the FIFPL fund covers part of the cost of Qualiopi-certified training under 2026 criteria set at €300 per day and €900 per year, with e-learning capped at 50% of the daily rate — funding that must be arranged before training starts, not after.
Frequently asked questions
Should a row with no known quantity be deleted from the DPGF?
No: it should stay visible and explicitly marked incomplete in the summary and in the À_CONFIRMER sheet, rather than deleted or filled with an unflagged estimate. Deleting a row makes a service that remains contractually owed disappear from the file. The quantity must be filled in as soon as it becomes known, without ever retroactively recalculating a total already sent to contractors without explicitly flagging the change.
Does the 10% VAT rate depend on the RGE certification of the contractor whose price is used as a reference?
No: that rate depends on the nature of the work and the dwelling's age, never on a contractor's RGE certification, even when their price served as the basis for pricing an item. Conflating the two distorts the applied rate without changing the actual scope of work. The rate used in the DPGF must therefore be checked independently of the source of the price used, item by item, before the document is circulated.
Can ChatGPT fill in a missing quantity from a similar earlier project?
No, unless explicitly requested and flagged as such: a quantity carried over from another project must be marked as an assumption to verify, never presented as a figure measured on the current project. Once the actual measurement is available, that assumption must be replaced and the corresponding row removed from the À_CONFIRMER sheet. Without that explicit instruction, the row must stay blank and flagged as missing data, just like any other unknown quantity.
What should be done if a significant gap appears between the DPGF and bids received?
Good practice is to review the quantity and price assumptions used in the DPGF, item by item, rather than mechanically adjusting the total to match the bids received. A systematic gap can also reveal a VAT error applied on one of the two documents. That review must be documented item by item in the register, to distinguish an explained gap from one that remains unexplained at the time of the final decision.
A reliable DPGF rests on three elements that do not change from one trade to the next: a checkable correspondence between every row and its CCTP article, a VAT rate confirmed item by item from the actual nature of the work, and incomplete items that stay visible until confirmed, never hidden behind a plausible value.
Once the DPGF is stable, our ChatGPT method for CCTP by trade details how to draft the technical requirements every item must answer, and our ChatGPT method for construction tender comparison revisits this same breakdown when comparing contractors consulted.
Training 100% fundable via OPCO/FIFPL. Qualiopi-certified programme. To structure your price breakdowns with AI as part of our AI for architectural management training, contact Educasium and specify your status (employee, self-employed, business owner) and your goal. This pack is also available on its dedicated ChatGPT skill page.