Buying software on an Indian research grant

Most lab software that never gets bought was never rejected. It stalled — in a queue for a committee that meets monthly, or against a request for three quotations nobody could produce, or on the question of which budget head a recurring cost belongs to. The decision was procedural, and it was made by somebody who never saw the software.

None of that is anyone's fault. Indian institutional purchasing was built for equipment — a thing, bought once, that arrives and can be inventoried. A subscription is none of those, and the rules bend awkwardly around it.

The number that decides everything

Before anything else, find out what your institution's delegation limit is for a direct purchase. Under the General Financial Rules, Rule 154 permits a direct purchase below ₹50,000 on a single competent signature, without quotations and without a committee. Above it, Rule 155 brings in a purchase committee and a heavier process.

Two cautions, both of which have cost people months:

• Your institution's own limit may be lower. The GFR figure is a ceiling, not a mandate; many institutes delegate less. Ask, in writing, before you plan around it.
• Establish whether the threshold is read inclusive or exclusive of GST. Different accounts offices rule differently and both readings are defensible. A price that clears it only pre-tax is a price that clears it in half of India.

Ask your accounts office those two questions first. Everything else in this guide is downstream of the answers.

Which head it sits under

A software subscription is almost always a recurring or contingency cost, not equipment. That matters more than it sounds: equipment heads often have money in them and recurring heads often do not, and moving between heads mid-project usually needs the funder's agreement.

Three things worth doing at proposal stage rather than afterwards:

1. Put the line in the proposal. A named recurring line — data storage, software subscription, data curation — is rarely contested by reviewers and is almost impossible to add later.
2. Ask for the full duration. A three-year subscription requested once is one approval; requested annually it is three, and the second one arrives when you are busy.
3. Say what it is for in the plan, not just the budget. A data management plan that describes careful curation with no budget line behind it is a plan relying on somebody doing it in the evenings. The plan guide covers that →

If the grant is already running and there is no line, contingency is usually the honest place to ask, and a small specific request approved quietly beats a large vague one referred upward.

What the finance office will ask you for

Almost none of this is about the software. Have it in hand before you start and the process takes days rather than weeks — most of the delay in a stalled purchase is one missing document at a time.

The three-quotations problem

Below the direct-purchase limit most institutions do not require quotations at all. Above it, or where local policy is stricter, you will be asked for three — and for software this is often genuinely impossible, because the products are not equivalent and some have no Indian reseller to quote at all.

What actually works, in order of how often it works:

• Get under the limit if the supplier has a plan that does. This is the whole reason some software is priced where it is.
• A proprietary article certificate. A short note from you stating that the item is available from a single source, with the reason. Most institutions have a form; it is routine, not exotic.
• Quote comparable alternatives yourself. Your office obtains competitor quotations — a supplier cannot supply them, and any supplier offering to is doing something you do not want attached to your purchase.

Ask your office which of the three they would prefer before you pick one. They have a route they like, and using it is faster than arguing for a better one.

Renewal, and the trap in a one-time order

A purchase order is an instrument for buying a thing once. A subscription renews, and the mismatch causes two failures that are worth planning around now.

Lapsing in the middle of a project

The order covers a year; the renewal falls due during a month when nobody is watching; access stops. Diary it eleven months out, not twelve, and ask the supplier to invoice before it expires rather than after.

Growing past what was approved

The lab grows and the plan no longer covers everyone. Ask before you buy whether additional people can be added to the same annual invoice at renewal, or whether each addition is a fresh purchase. The answer changes how you write the first order.

And ask the exit question while you are still a prospect, because that is the only time you will get a plain answer: if we stop paying, what happens to what we put in? An acceptable answer names a format and says export is free. An unacceptable one talks about data retention policy.

If the answer comes back no

Sometimes it is a genuine no — no money in the head, no policy route. More often it is a no to how it was asked. Three things worth trying before accepting it:

• Ask what would make it a yes. Purchase officers almost always know, and are rarely asked.
• Ask a colleague who has bought software recently. Within one institution the successful route is usually a specific person and a specific form, and it is not written down anywhere.
• Check whether the institution already has it. Central IT licences things that individual labs then buy again. It happens more than anyone admits.

And if it is a real no, find out whether a free tier does enough to keep the record from decaying while you wait for the next funding cycle. A worse tool used consistently beats a better one bought in eighteen months.