The India Architect Fee Benchmark
What do Indian practices actually charge? Nobody has published an honest answer, so we are collecting one. The method, the full question set and the privacy rules, stated in advance. Fieldwork closes 30 November 2026; results publish January 2027.
- Nobody in Indian practice knows what anybody else charges. Fee conversations run on anecdote, and the anecdote is usually the lowest number somebody heard.
- This benchmark collects real fee data from Indian practices and publishes it as medians and ranges, by project type, city tier and studio size.
- No studio is named. No cell with fewer than five responding practices is published at all.
- Contributors get the full dataset breakdown before it is published, and can withdraw at any point up to publication.
Ask ten Indian architects what a full design service on a private house should cost and you will get ten answers, four of which are the Council of Architecture minimum recited from memory, three of which are what the last client agreed to, and three of which are a number somebody heard at a conference. There is no published, methodologically honest picture of what practices in this country actually charge.
That absence is not neutral. It costs money, and it costs it in one direction. A principal with no reference point negotiates against a client who has rung four studios and does have one. Every fee conversation in the country is asymmetric, and the asymmetry runs against the profession.
Why an existing number is not enough
Two figures already circulate, and both are doing a different job from the one a practice needs.
| Source | What it tells you | What it cannot tell you |
|---|---|---|
| The CoA scale of charges | A published minimum by category: 7.5% for an individual residence, 5% for most non-housing work. See what to charge. | What anybody actually charges. It is a floor, not a distribution, and it has no view on city, studio size or project type beyond its categories. |
| Rate cards and forum threads | What individual studios say in public, which is a real signal. | Whether it is representative. Practices that publish a rate card are a self-selected and unusually confident minority. |
Neither is wrong. Both are answering a question other than "where do I sit relative to practices like mine", which is the question that actually changes what you quote on Monday.
A minimum tells you the floor you must not go below. A benchmark tells you the room you did not know you had.
What this will report
The structure below is fixed and is what contributors are being asked to agree to. The cells are empty because the fieldwork is still open, and they will be filled with what the responses say rather than with what anyone expected them to say.
| Project type | Median % of cost | Interquartile range | Median ₹/sq ft | n |
|---|---|---|---|---|
| Individual residence | — | — | — | — |
| Apartment interiors | — | — | — | — |
| Retail and F&B | — | — | — | — |
| Commercial / office | — | — | — | — |
| Institutional | — | — | — | — |
| Group housing | — | — | — | — |
Medians and interquartile ranges, never means: one studio quoting an outlier moves a mean and does not move a median. Any row with fewer than five responding studios will be suppressed rather than published thin.
| Cut | Median % of cost | Median ₹/sq ft | n |
|---|---|---|---|
| Metro (Mumbai, Delhi NCR, Bengaluru, Hyderabad, Chennai, Pune, Kolkata) | — | — | — |
| Tier 2 | — | — | — |
| Tier 3 and smaller | — | — | — |
| Sole practitioner | — | — | — |
| 2 to 5 people | — | — | — |
| 6 to 15 people | — | — | — |
| 16 and above | — | — | — |
Three further things will be reported alongside the fee tables, because they are the questions that come up immediately after "what should I charge" and nobody has numbers for them either: how long practices actually wait to be paid, measured in days from stage invoice to receipt; how many revision rounds studios include and how many they end up absorbing; and what proportion take an advance before drawing anything.
The method, stated in advance
A benchmark is only worth as much as its method, and a method disclosed after the results are known is not a method. Here is the whole of it.
| Rule | |
|---|---|
| Who is eligible | Practices carrying out architectural work in India, of any size, whether or not the respondent is CoA-registered. One response per practice; where two people from one studio respond, the later is discarded. |
| Fieldwork window | Opens September 2026, closes 30 November 2026. Fees are collected for work quoted in the calendar year 2026, so the dataset has one time base rather than a mixture. |
| How it is collected | A structured questionnaire, submitted through the form on this site. Free-text answers are coded by hand; nothing is inferred from a studio's public rate card. |
| What is reported | Median and interquartile range. No means, because a single outlier moves one and not the other. n is stated on every table. |
| Minimum cell size | Five. A cut with fewer than five responding practices is suppressed entirely rather than published with a caveat, because a caveat below a table does not stop the number being quoted. |
| Outliers | Retained, not trimmed. A median is already robust to them, and removing inconvenient responses is the thing that makes a survey worthless. |
| Anonymity | No practice is named, in the publication or anywhere else. City is reported as a tier, never as a city, for any cut where naming it would identify a respondent. |
| Raw data | Never published, never sold, never shared with a third party. Held only for as long as the edition it feeds, then deleted. |
| Withdrawal | Any respondent may withdraw their response at any point up to publication, by replying to the acknowledgement email. No reason required. |
| Who is paying for it | Nobody. It is run by AtelierLab, which makes practice software, and is published free. No sponsor has a say in the questions or the findings, because there is no sponsor. |
If any of the above changes before publication, the change will be stated on this page rather than made quietly.
What you will be asked
Twelve minutes, and no question requires you to look anything up. The full instrument, so nobody is agreeing to something unseen:
- About the practice. City, size in people, years in operation, and the two or three project types that make up most of your work. No practice name is required and it will not be published if you give one.
- How you price. Percentage of cost, per square foot, lump sum, or a mixture, and which you use for which project type.
- The numbers. For your most recent completed commission in each of your main project types: the fee as quoted, the basis, the built-up area, and the approximate cost of works.
- What sits on top. Whether you charge the 10% documentation charge, and whether GST is quoted inclusive or on top.
- Stages and payment. How you split the fee, the advance you take, and the typical number of days from a stage invoice to the money arriving.
- Revisions. Rounds included per stage, rounds typically absorbed, and whether you have an hourly rate stated in the appointment letter.
- One free-text question. The fee decision you would make differently if you were starting the practice again. This is the answer most likely to be worth reading, and it will be published as anonymised quotations only with explicit permission.
What you get for taking part
- The full breakdown before it publishes, including the cuts that will be suppressed publicly for being too thin, which are frequently the ones nearest to your own situation.
- Your own position in the distribution. Which quartile your fees sit in, for your project type, your city tier and your studio size. That single number is the thing most principals have never had.
- No sales call. Taking part puts you on the list for the results and nothing else. If you want to hear about the software you have to ask, and the form has a box for that which you are free to leave alone.
The reason to be candid is straightforward: a benchmark built from thirty studios understating their fees produces a number that then gets quoted at all thirty of them by their own clients. This is one of the few professional exercises where accuracy and self-interest point the same way.
While you wait for the results
The benchmark will tell you where you sit. It will not tell you how to make the number stick, and those are different problems with different remedies.
- What to charge, and how to build the number — the CoA floor, the three pricing models, and the five appointment-letter clauses that make a fee hold.
- The stage-wise schedule, and the contract around it — where the money should arrive, and the eleven clauses that decide whether it does.
- Scope creep and revision control — the twelve hours a “small” revision actually costs, and the log that makes it visible.
Take part in the benchmark
Twelve minutes, closes 30 November 2026, results to every contributor before publication. Choose “Fee benchmark” on the contact form and we will send you the questionnaire. If you would rather see the practice software this is published by, that is here — but the benchmark asks nothing of you in return.
Questions, answered
The questions architects ask most about this, in plain language.
Ask us anythingFieldwork closes on 30 November 2026 and the first edition publishes on this page in January 2027. Contributing practices receive the full breakdown, including the cuts that will be suppressed publicly for being too thin, before it goes up. If those dates move, the change will be stated on this page rather than made quietly.
No. No practice is named in the publication or anywhere else, and no practice name is required to take part. City is reported as a tier rather than as a city wherever naming it could identify a respondent, and any cut with fewer than five responding practices is suppressed entirely rather than published with a caveat — because a caveat under a table does not stop the number being quoted.
It is never published, never sold and never shared with a third party. It is held for as long as the edition it feeds and then deleted. You can withdraw your response at any point up to publication by replying to the acknowledgement email, with no reason required.
Because one studio quoting an unusual number moves an average and does not move a median. Fee distributions in a small profession are skewed — a handful of practices charge multiples of the rest — and a mean of thirty responses would describe none of them. Medians and interquartile ranges are reported instead, with n stated on every table, and outliers are retained rather than trimmed, since removing inconvenient responses is the thing that makes a survey worthless.
Nobody. It is run and published free by AtelierLab, which makes client-management software for architecture practices. There is no sponsor, so no sponsor has a view on the questions or the findings. Taking part puts you on the list for the results and nothing else; if you want to hear about the software you have to ask for it.
No, and they answer different questions. The CoA scale is a published minimum by category of work — 7.5% of the cost of works for an individual residence, 5% for most non-housing projects — and it is a floor rather than a description of the market. This benchmark measures what practices actually charge, and reports the distribution around it. A minimum tells you what you must not go below; a benchmark tells you the room you did not know you had.
Particularly so. Sole practitioners and studios of two to five are the largest part of Indian practice and the part with the least published information about it, which means it is also the part most exposed in a fee negotiation. Studio size is one of the cuts the benchmark reports, so the more small practices respond, the more likely it is that the cut nearest to your own situation clears the minimum cell size and gets published at all.