Pkeday.

Mainboard · RHP filed 2026-10-08

Fusion CX Limited

KILL First look 12 Oct 2026 · process v3.2-sandbox

Stopped at the first look. The best of its seven lines scores 3 out of 5 on its market and 3 out of 5 on its right to win, and the growth that makes the company look fast was bought: setting aside the firms it acquired, revenue fell about 8% last year, and two-thirds of what was left came from one customer that also arrived with an acquisition.

Rs 275–289 a share · bids 14–16 October 2026 · Rs 702 crore raise: Rs 500 crore fresh, Rs 202 crore sold by the two promoter companies. The band is not printed in the document and comes from the National Stock Exchange alone.

Fusion CX runs outsourced customer support for large American companies — calls, chats, emails, order handling and collections — from 40 centres in 13 countries, and has grown mostly by buying other firms that do the same thing.

The bearings

Market value at the top of the band
About Rs 4,258 crore
Price against last year's profit
About 21.5 times. The filing's own peer list runs from 14 to 51 times and averages 27, so the price is in the lower half of it
Fresh money against shares sold by holders
Rs 500 crore fresh (71%), Rs 202 crore sold by the two promoter companies (29%). Each sells only about a twentieth of what it holds, at a cost of 12 to 13 paise a share. The two individual promoters hold no shares directly and sell nothing
What the fresh money does
Rs 275.7 crore repays debt, which is 86% of all borrowings. Rs 61.1 crore goes into its own software platforms, the only business line the money reaches. The rest is unsized, for acquisitions it has not named and general purposes
Promoter holding
96.21% before the offer, about 80% after. There has never been an outside shareholder
Last private price
No outside investor has ever priced these shares. The only priced dealings are employee options at Rs 60 and Rs 62 in September 2026 and a 100-share transfer at Rs 72.20 in January 2025, so there is no market price to compare the band with
Latest full year
Revenue Rs 1,818 crore, up 37%. Profit Rs 170 crore, up 129%. Setting aside the businesses it bought, revenue fell about 8%
How the offer is split
The company does not meet the profit record the main eligibility route requires, so it is using the alternative: at least 75% of the offer goes to institutions and at most 10% to retail buyers
Lead managers
Nuvama Wealth Management, IIFL Capital Services and Motilal Oswal Investment Advisors

What the company sells

Customer service

41% of revenue · +36% last year

Answering phone calls, chats and emails on behalf of large American companies, mostly telecom and power utilities, from the company's own centres and staff.

The market3/5

Nobody publishes a growth rate for the size of firm this is, so one was built from eleven listed competitors: it comes out at about 8% to 12% a year. Some of those competitors are shrinking — one of the two largest guides its traditional work down 3% to 5% this year on faster automation, and names USD 1.3 billion of work being squeezed. The company's own paid-for research, read carefully, implies the non-digital part of this market shrinks slightly every year, and that is where most of this revenue sits. The score is the market's own pace and nothing more: there is no sign of this company gaining ground in it.

Right to win3/5

One real and unusual advantage: the company is a certified minority-owned business in the United States, with five certificates from named bodies, held since 2015. That matters because a California regulation sets utilities and telecom carriers there a 15% minority-supplier purchasing target, checked through the same clearing house — and telecom and utilities are 47% of this company's revenue in a country that is 85% of it. No large competitor can copy it at any price, because the rule requires majority ownership by qualifying individuals. What holds the score down is that nowhere in 559 pages is a single rupee of revenue, a contract or a customer attributed to that status, and the filing never says whether it survives the listing. The two things the company does point to both fall apart: the 13-year customer relationships are bought assets, recorded in the accounts as purchased customer lists, and the same American telecom contract has changed hands four times in a decade, selling for about USD 146 million in 2016 and about USD 12 million in 2025. Eight of its ten largest customers were lost over three years. And the 28 languages from 40 centres are matched or beaten by name — one competitor its own size publishes the same 13 countries and more languages.

Winning new customers for clients

26% of revenue · +32% last year

Outbound sales calls, finding sales leads and running sales campaigns for the client, sold to the client's commercial side rather than its service desk.

The market3/5

Outsourced sales work bought from the United States is about USD 6.6 billion and grows about 10% a year, and there is a real shift behind it: the work is moving from American agents to Latin America, which the company is riding — its own onshore share of revenue fell from 50% to 32% in nine quarters while its margin rose. Three to five years of that remain. Against it, scripted outbound calling is among the most automatable work there is, a 2024 American ruling restricts artificial voices on calls, and the company's own risk factors concede clients bring this work back in-house. Its own share of revenue fell from 32% to 26% in two years, which is mostly because the businesses it bought do other things — but on the June quarter's pace this line is growing 2%, against a market growing 10%, so it is losing ground.

Right to win2/5

Nothing here is hard to copy. The global network of centres and languages is matched by name by competitors of the same size and beaten by the larger ones. The long customer relationships are the same bought assets as above. A calling tool acquired for this line has no disclosed result of any kind. The only outside evidence is sixteen customer reviews averaging 4.6, most of them from 2017 to 2022 and now too old to rely on. And the deeper problem is how the work is paid for: across the company, work priced per person per month has risen to 91% of revenue while work paid for by results has halved to 6% — and fallen in rupees. Sales work is the kind most naturally paid for by results, so a company doing it for a headcount fee is being paid for labour rather than for selling.

Order handling support

15% of revenue · +57% last year

Taking and tracking orders, arranging installations, fixing failed orders, and handling returns and refunds for telecom, utility and retail clients.

The market2/5

No one publishes a size or a growth rate for this work at all. Reading across the listed competitors it is somewhere between 3% and 8% a year, with a central estimate of 5%: one large competitor guides this kind of work down on faster automation, another's trust-and-safety work fell 12%. There is a genuine driver — America's fibre rollout generates orders for three to five years — but the company is paid per person per month, not per order, so more orders do not automatically mean more revenue. And last year's 57% is not shown to be the company's own growth: this line fell to 11% of revenue in the June quarter, down 16% on its own run rate.

Right to winnot scored

Keeping clients' customers from leaving

5% of revenue · +111% last year

Calling customers who are about to cancel, winning back those who have left, and processing renewals inside the client's own systems.

The market2/5

Nobody sizes this market — not the company's paid-for research, and not any independent analyst we could find after a proper search. That silence is itself the answer at this stage. The nearest guide is the group of mid-sized listed competitors, growing 10% to 29%. The line grew six times in two years, but from a base of about USD 1.8 million, and last year's increase of about Rs 49 crore sits against Rs 135 crore of revenue bought into the same three industries by two acquisitions — so the growth is most likely bought, not won. It is also slowing, to about 28% on the June quarter's pace.

Right to win2/5

There is one specific and genuinely useful thing here: the company sits inside the client's own renewal and payment approval process, supplying the tools, the telephony and the reporting. That is a real reason for a client to keep it. But it has no disclosed result attached to it anywhere. Everything that would make this a strong business belongs to the client — the model that predicts who will leave, the data on who is eligible for what, the offer made to them and the customer records. And there is no save rate anywhere in 559 pages: a search of the whole document turns up one case study with no number in it. Four named competitors publish exactly this figure, one of them a 50% fall in cancellations and tripled saves, so the absence here counts against the company rather than being neutral.

Technical support

4% of revenue · +51% last year

First-line help desk for products and devices — the first person a customer reaches when something does not work.

The market2/5

First-line help desk work is going nowhere: built from the listed competitors, this work runs from minus 5% to plus 3% a year, with a central estimate of flat. The broader pool above it grows 5% to 6%, but that is not the work this line does. It is also the work most directly replaced by automated answers. There is no reliable sign of the company gaining ground, and the score is the market's own pace.

Right to winnot scored

Its own software (Arya and MindVoice)

size not disclosed · not disclosed

Software sold to contact centres: tools that check call quality and prompt agents while they are talking to a customer.

The market2/5

The market itself is decent — software of this kind grows about 20% a year. The company's position in it is the problem, and the filing settles it. The only revenue figure for this software anywhere in 559 pages is a part-year stub from two years ago, about Rs 18 crore annualised, with a loss of about Rs 5 crore, and nothing at all for the two years since. The audited accounts record all revenue as services and none of it as the kind recognised over time, so there is no subscription business. The people are going the wrong way: development staff fell from 74 to 36, development spending is 1.27% of revenue and falling, and 36 people work on it in total, four of them in sales. One of the two products is powered by another company's technology, which contradicts the claim elsewhere that it is built in-house. A quarter of the Rs 61 crore pays salaries already being paid and over half pays cloud and licence bills. The company bought this software from its own promoters for USD 10,000 in December 2023.

Right to winnot scored

AI data work

size not disclosed · not disclosed

Labelling images and video to train artificial intelligence, driving robots remotely, and filming demonstration footage, delivered from India, the Philippines and Morocco to American buyers.

The market3/5

This market is real and fast: roughly Rs 8,000 to 17,000 crore growing 23% to 33% a year, and that was confirmed outside the company's own research by a listed competitor whose equivalent business has been its fastest-growing line for six straight quarters. The company's position earns nothing, because nothing is disclosed: no revenue, no customer count, no capacity. What keeps the score at the market's pace rather than lower is that this looks like a genuine start rather than an empty claim — one of the two brands launched in July 2026, the other was hiring 400 people in India in September, and two of the four customers named on the company's own websites also appear inside the prospectus.

Right to win2/5

The things the company brings are real but not hard to copy: cheap hiring at scale, 40 centres, and good security and health-data credentials. The trouble is that everyone who competes here already has them. One listed competitor's equivalent business is on its own bigger than all of Fusion CX. Another large firm agreed in June 2026 to buy an Indian competitor in this exact business for up to USD 310 million — and what it bought was the software platform and the expert network, not the delivery centres. That competitor already works from the same two Indian cities. A company founded this year runs the remote-driving work across three countries with no buildings at all, and training a remote operator takes between eight and forty hours. Prices are falling hard: remote driving went from about USD 340 an hour in early 2024 to about USD 120 to 135 in early 2026, while the work is still sold by the month per person.

Also on the books and not scored: billing and collections (6% of revenue, up 9% last year) and digital solutions (5%, up 4% after a one-off jump the year before). Neither is a story on the year.

The numbers that matter

MeasureFY24FY25FY26Jun 2026 qtrReads
Revenue (Rs crore)9911,3291,818490Up 37% last year. Setting aside the businesses it bought, revenue fell about 8%
Largest customer, share of revenue14.4%17.7%30.1%25.5%That one customer's growth was about two-thirds of the company's whole increase last year
Ten largest customers, share of revenue41.0%—52.2%49.8%Concentration has risen sharply in two years
Clients paying over USD 1 million a year—3733—Fell in the year revenue rose 37%
Work priced per person per month81.0%—89.2%91.3%Rising. Work paid for by results halved to 6.25%, and fell in rupees too
Staff leaving each year32.2%30.6%26.5%31.2%Better last year, then back up in the June quarter
Revenue per employee (Rs lakh)11.610.613.614.3Among the lowest of its listed competitors, which is what buying cheap labour looks like
Goodwill and similar assets, against the company's own capital———75%Rs 464 crore, four times its level two years ago, and never written down

Checked in the filing

What happens next

The full assessment stops here. If you would like it run anyway, say so and it will be added to this page.

Read from the red herring prospectus of 8 October 2026 (559 pages, by section), with outside research on the markets, the rivals and the acquisitions. Market sizes the company paid a consultant for were checked against listed competitors' own filings and results. The split between bought and underlying growth was rebuilt from the prospectus's own notes on the businesses it acquired, and the history of one acquired contract was traced through another company's filings with the American regulator.