Decisions that cost orders before they freed cash

Rewriting the milestones after ₹520 crore went unbilled

When I reviewed the Projects division's balance sheet in early 2020, it held ₹520 crore of contract assets against annual revenue of ₹980 crore. Most of our material handling and effluent treatment contracts paid 10 per cent as an advance, 60 per cent on delivery of equipment to site and the last 30 per cent on mechanical completion and performance tests. We recognised revenue as costs were incurred, which was correct. But completion depended on clients' own civil works and power supply, and on four large sites those had slipped by more than a year. ₹164 crore of our contract assets were over a year old, and we were financing our clients' delays at our own borrowing cost of about 10 per cent.

I proposed three terms for new bids: erection billed monthly on measured progress, deemed completion if a client's delay exceeded ninety days, and no more than a fifth of value payable after mechanical completion without the bid committee's explicit approval. Prakash Odedra, president of Projects, opposed them. Steel-plant clients write their own terms, he argued, and award work to whoever accepts them. He was right about the next two tenders. We lost a ₹380 crore conveyor package and a ₹260 crore raw-material handling system to competitors who accepted back-ended schedules.

We also cleaned up the past. We negotiated partial billing on three of the four delayed sites, and on the fourth we booked a ₹38 crore expected-credit-loss allowance and later settled for less than we were owed. That charge landed in 2020–21, a pandemic year in which group revenue had already fallen to ₹1,870 crore.

By March 2026 Projects contract assets were ₹270 crore on revenue of ₹1,320 crore, and balances older than a year had fallen to ₹29 crore. Two clients who rejected our terms in 2020 accepted them on later orders. Every Projects bid now reaches committee with a cash curve, and the division's managers are measured on operating cash flow as well as margin.

I would do two things differently. I would have separated our claims for client-caused delay from the argument about future terms, and pursued them earlier with our own site records; we had the evidence and used it late. And I would have shown Prakash the cash curve of his own best contract before arguing about anyone else's.

Locking plate and rupee on the day an order is signed

In 2019 our Process Equipment division bought plate when engineering released fabrication drawings, usually three to five months after an order was signed, and treasury covered currency against an annual budget rate on a rolling half of forecast receipts. Between signing and buying, the division carried an open position in steel and, on export orders, in dollars and euros. Over the three years to 2019–20, the margin delivered on fixed-price orders had fallen short of the margin at bid by 4.6 percentage points on average, mostly because of plate.

In September 2020 I proposed that, for every fixed-price order above ₹25 crore, procurement reserve plate with mills at signing under price-locked agreements, and treasury book forward contracts matched to the order's milestone dates. Nilesh Jariwala, our head of procurement, disagreed. Buying across several orders at once earned discounts of 2 to 3 per cent, and he could time purchases into dips. Our managing director worried about the advances mills would demand. Both concerns were fair. We kept pooled buying for smaller orders and accepted the loss of about ₹14 crore a year in volume discounts, with average advances to mills of about ₹110 crore.

The test came quickly. Plate prices rose by about half over the following year. Fixed-price orders executed in 2021–22 and 2022–23 delivered margins within 0.8 percentage points of bid on average, and we never had to ask an export client for relief. The forwards, designated as cash-flow hedges under Ind AS 109, kept currency movements inside the bid assumption.

The policy also cost us. When prices eased in 2022–23, our locked plate cost ₹27 crore more than spot purchases would have. One mill under financial strain held ₹18 crore of our advances for seven months before delivering. Nilesh was entitled to remind me of both, and he did.

Every bid paper now states the plate tonnage, the locked price and the hedge cover before the bid committee approves it, and procurement and treasury report open positions weekly.

If I did it again, I would cap advances to any single mill from the start. I would also negotiate harder for price-variation clauses with public-sector clients, instead of carrying domestic risk that the client could have shared at little real cost to itself.

Recovering ₹410 crore of bank limits instead of pledging the chairman's name

By April 2021 Ilvara had used 92 per cent of the ₹1,900 crore of non-fund-based limits sanctioned by its eight consortium banks, and the sales team had ₹2,100 crore of bids in preparation that would need about ₹420 crore of new guarantees if we won. Our executive chairman, Mihir Rajyaguru, proposed asking for ₹800 crore more, offering promoter-group land as collateral and his own personal surety. The lead bank was willing. It was the quickest answer, and a generous one.

I recommended against it. Limits secured on the family's assets would deepen the habit that had filled the lines, tie the promoter group's property to the company's contract risk and, I expected, weigh on how rating analysts read our contingent liabilities. My alternative was slower. When we examined every outstanding instrument, ₹340 crore related to contracts whose defect-liability periods had ended, but clients had not returned the originals, so the banks still counted them. A team of four began recovering them client by client, with formal notices where courtesy failed. We asked clients to halve performance guarantees after successful tests, which some contracts allowed. And we withdrew from three bids worth ₹760 crore whose bank requirements were heavy and margins thin.

The cost showed in 2021–22, when order intake fell 9 per cent and our sales team watched competitors take work we had let go. Mihir accepted the plan without pretending to like it, and the board asked for monthly progress.

Within fourteen months we had released ₹410 crore of limits, ₹290 crore through returned instruments and ₹120 crore through reductions, and utilisation fell to 68 per cent. Our rating rose from BBB+ to A- in September 2022 and to A in November 2024, and the commission we pay fell from about 1.4 per cent a year to 0.9 per cent. In 2024 the lenders released Mihir's personal surety on our term loans, and in 2025 they raised our limits to ₹2,600 crore without promoter collateral.

What I would change is what allowed the crisis. We should have kept a live register of every instrument's obligations long before the lines filled. And I would have shown sales the bank-capacity cost of each bid from the start, not after we had withdrawn from three.

Taking ₹21 crore out of the EBITDA we were selling

In April 2018 Tamarind Peak Capital began preparing to sell its 46 per cent of Varunika Castings, together with most of the founder's holding. The fund's advisers drafted an information memorandum showing adjusted EBITDA of ₹151 crore for 2017–18, against reported EBITDA of ₹142 crore, by adding back ₹9 crore of start-up losses on a new moulding line. As CFO I was responsible for the numbers behind that document, and I did not believe ₹151 crore would survive a buyer's diligence.

Two items worried me more than the add-back. ₹12 crore of our EBITDA came from export incentive income under a government scheme whose future was then uncertain, so a buyer could not rely on it. And ₹9 crore of pattern and tooling costs had been capitalised on the expectation that customers would reimburse them, which most had not. That policy had been adopted in 2015 on my own recommendation. I proposed presenting ₹130 crore as the sustainable figure, with reasons, and disclosing the start-up losses rather than adding them back.

Kunal Hegdekatti, the fund's partner on our board, disagreed sharply. At the multiples being discussed, ₹21 crore of EBITDA was worth around ₹200 crore of value, much of it to his investors. The advisers said every seller presents adjusted figures. Shrikant Mohite, the founder, asked me only whether I could defend each rupee. I told him I could defend ₹130 crore and not ₹151 crore.

We went to market with ₹130 crore, and two of five bidders withdrew. The winning buyer, a listed Indian auto-components group, paid an enterprise value of ₹1,240 crore, about 9.5 times our figure, against indicative bids that had opened near ₹1,400 crore. Its confirmatory diligence took six weeks instead of the four months the advisers had planned, and in the two years after completion it made no warranty claim. Kunal later told his investment committee that the clean diligence had been worth part of the discount.

What I would do differently starts with my own policy. I should have reversed the tooling capitalisation in 2016, when reimbursements first fell short, rather than leaving it for a sale to expose. And I would commission a vendor's quality-of-earnings review two years before any exit, while there is still time to fix what it finds.