Turnover Criteria That Exclude Good Vendors

Opinion, so labelled. A ministry asked for average annual turnover of fifty crore for an on-prem agent that would sit on a single VLAN and draft file notes. The firms that cleared the gate were system integrators who would subcontract the actual model work to a twelve-person studio. The studio was the only party who understood the air-gap. The studio could not bid.

That is not prudence. That is a filter imported from road contracts, where turnover is a rough proxy for dumpers and cash. Agentic software does not have dumpers. It has architecture. GFR already knows this well enough to let recognised startups skip prior turnover. Many ATCs have not caught up.

This is not legal advice. If you are a buyer, ask counsel before you change a standing template. If you are a vendor, do not shout that a clause is illegal because it hurts.

What turnover was for

Qualification exists so that a winner can survive the job. In works, a firm with no revenue history often cannot pay labour through a delayed RA bill. In commodity supply, a hollow company can vanish after EMD. Those fears are not stupid.

They become stupid when the article is a specialised system whose failure mode is egress, not insolvency, and when the bid already takes performance security. A profitable reseller with a rented team can pass a turnover test and still ship prompts to a foreign host. A DPIIT-recognised firm with a clean on-prem product can fail the test and never be scored.

What GFR already allows you to do

Startup India points at Rule 173(i) for relaxation of prior experience and prior turnover for DPIIT-recognised startups, subject to quality and technical specifications. The same ecosystem points at Rule 170 for EMD. The Manuals for consultancy and for works have parallel sentences. The Union has been saying, for years, that new firms should not be excluded by history alone.

A buyer who copies a five-times-estimate turnover line from a furniture bid, and then forgets the startup clause, is not following a hidden third rule. They are on autopilot.

Risk you think you are buying versus risk you actually buy

Gate	Fear it addresses	What it misses in an agent buy	Better proxy
Average annual turnover	Hollow company, no cash	A rich SI with a leaking architecture	Performance security + escrow of install media
Five similar works	Has done this before	Has done chatbots, not air-gaps	A witnessed install on your sample data
ISO binder	Has a quality system	The certificate never names the model host	A data-flow diagram you can test
DPIIT waiver, used well	Lets new firms in	Still needs a technical fail line	Equal tests for everyone

Two files, two outcomes

References are often the same gate in a hat

When turnover becomes embarrassing, some ATCs switch to three completed government AI projects of similar value. For a first-generation on-prem agent, that sentence is a closed shop. The first government buyer you want cannot exist if every buyer demands a predecessor.

If you need comfort that the firm can deliver, ask for a witnessed install on sample data, a runbook, and named engineers who will be on the VLAN. Those tests work on a two-year-old company. A reference from a ministry that bought a hosted chatbot in 2023 does not tell you whether this stack leaks. It tells you the firm has a logo.

Opinion, again: a department that wants sovereign capability should stop selecting for the ability to have already sold to government. That selection is how the same three integrators appear on every file, wrapping the same six studios. The studios never learn to contract. The integrators never learn the model. The perimeter stays a slide.

A turnover multiple also interacts badly with GeM transaction charges and delayed CRAC. You asked for a fat P&L so the winner could float. Then you paid late. You selected for a prime and then used that prime as a bank. If float is the real fear, write milestone invoices and a named CRAC owner. That is cheaper than excluding every firm that actually built the product.

The opinion, tightly

For sovereign agent platforms, turnover should be a residual tool, not the first page. Prefer: a technical compliance matrix that can fail a hyperscale brand, a security schedule, a training ban, an exit test, and a performance security sized to the work. If you keep turnover, scale it to mobilisation need, not to a multiple you inherited from a PWD manual. If you keep it, waive it for recognised startups and say so in the same table.

Do not replace turnover with a demand for three government references if you are the first government buyer you want. That is the same exclusion with a different hat.

Objections from risk-averse rooms

Audit will ask why we let a small firm in. Audit will also ask why you paid a reseller 40 percent to borrow a product. Write the technical tests. That is the answer to both.

Startups fail. So do integrators. Performance security, staged go-lives and a right to terminate for security breach are how grown-up contracts handle failure.

GFR makes us put a number. GFR makes you think. It does not make you paste five-times-estimate onto a model runtime.

What a better PQ table looks like

Put technical fails above money history. Deployment must be on infrastructure the institution controls. Training on customer data must be off unless a written exception exists. Logs and plan traces must export in a named format. Exit must include a deletion certificate that names backups. A bidder who fails any of those is non-responsive, whether their turnover is two crore or two thousand.

Then, if you still want a financial gate, size it to mobilisation. How many months of staff will the winner float before the first CRAC. That number is a reason. Five times estimate copied from a works manual is a superstition. Superstitions exclude the firms the startup policy was written to include, and they do not exclude the reseller who can rent a balance sheet.

Write the DPIIT waiver in the same table, not in a footnote. Turnover: X for others; waived for recognised startups subject to the technical fails. Experience: Y years or Z similar works for others; waived for recognised startups subject to a witnessed install. EMD: as per Rule 170. Officers can evaluate a table. They cannot evaluate a vibe that the relaxation might apply.

If a standing template will not let you edit those cells, put a one-page deviation note on the file and ask the competent authority to approve the deviation for this category of buy. Templates are how last decade's furniture bid governs this decade's model runtime. Competent authorities exist to break that spell on purpose, in writing.

Playbook if you write ATCs

• This month: find the standing turnover sentence in your template.
• Ask finance whether Rule 173(i) has been considered for this category.
• Rewrite the PQ table: technical tests first, turnover as optional, startup waiver explicit.
• Run one bid that way. Keep the evaluation report.

Note for the indent file

The estimate is for a specialised on-prem agent. A high turnover multiple would likely route the work through a reseller. It is proposed to rely on technical tests, performance security and, for DPIIT-recognised startups, the relaxation from prior turnover under the GFR / Startup India framework. Quality specifications are not relaxed. This is a policy choice for competent authority, not a vendor request copied onto the file.

Prcept AI is a DPIIT-recognised firm. We would rather lose a scored bid than win a world where only wrappers can enter. If your risk is architecture, test architecture.

How a buyer or seller should act on this

Treat “Turnover Criteria That Exclude Good Vendors” as an operating problem, not a thought piece. A P5 Founder who searches “turnover criteria government tender” is usually one bid, one CRAC, or one rejection away from a cash event. The file that wins is the one with dates, document names and a named officer — not a paragraph that restates GeM’s homepage.

A five-times-estimate turnover gate is a risk habit from civil works. For novel agents it often buys you a reseller, not a safer stack. This is an opinion, labelled as one. That is why this opinion ends in artefacts: screenshots of the live portal term, the clause you invoked, and the date you last checked it. GeM, GFR notes and state portals move. A citation without a date is folklore.

Confirm the live GeM FAQ, the current revenue policy and the bid text before you copy any number from a blog — including this one. If the portal and this article disagree, the portal wins. Put the printout in the file.

• Write the purpose of the buy in one sentence a DDO will sign.
• Name the route: catalogue, custom bid, bid, RA, CPPP, or state portal.
• Attach the exemption or preference documents you will actually upload (Udyam, DPIIT, MII, OEM).
• Record who can accept the consignee receipt and who raises the bill.
• Do not invent a category, a PAC, or a price-reasonableness story after L1 is public.

Close this loop before the next CAB

Put “Turnover Criteria That Exclude Good Vendors” on the next change-advisory or bid-opening agenda as a single line item with an owner. If it cannot earn a line item, it will not earn a control. The owner should be a P5 Founder, not “the vendor.”

Revisit the item when the model, the GeM term, the region, or the SI changes. “turnover criteria government tender” is not a one-time workshop. It is a watch item. Date the last check. Unsigned watch items are souvenirs.