A first-year analyst at a mid-market Indian investment bank spent three days building a DCF model for a packaging company acquisition. The model was technically correct. Every formula linked properly. The sensitivity tables populated cleanly. When he presented it to the VP, she looked at it for four minutes and said: "Your revenue assumptions don't match the capex trajectory. A packaging business growing revenue 18% annually would need a greenfield plant in year three. Where is that in your model?"
The analyst had built a financially consistent model that was operationally incoherent.
This is the gap that separates technically trained financial modellers from investment banking analysts who are actually useful on a deal. The mechanics are learnable in three months. The judgment — understanding the business well enough to know whether the numbers make operational sense — is what takes longer and what separates the analysts who get staffed on the consequential mandates from those who stay in the data room.
Why 2026 Demands More Than Template Proficiency
The analyst skill set expected at Indian and global investment banks has shifted materially in the last three years. Three forces are driving this:
AI-assisted modelling is raising the floor. Tools like Microsoft Copilot, custom GPT models integrated into Excel, and Bloomberg's AI-enhanced terminal mean that assembling a basic three-statement model is increasingly automated. A first-year analyst who can only build what a template generates is competing with software. The value shifts to interpretation, assumption-setting, and knowing when the model is wrong.
Deal complexity is increasing. The Indian M&A market in FY2024 saw transactions involving cross-border structures, earn-out provisions, contingent consideration, and multi-currency financing that did not exist in simpler deal structures five years ago. Models that cannot handle these mechanics are useless on the mandates that matter.
Turnaround time is compressing. Pitchbooks that once took a week now need to go out in 48 hours. Analysts who cannot build and audit a model quickly — because they are still learning the mechanics — consume time that senior bankers do not have.
The implication: technical fluency is no longer the differentiator. It is the admission ticket. The differentiator is the combination of technical fluency, business judgment, and the ability to communicate what the model is actually saying.
The Three-Statement Model: Where Judgment Separates Analysts
Every IB analyst can link a three-statement model. What separates them is whether the model is economically coherent — whether the operating assumptions imply a business reality that makes sense.
The non-obvious skill: reverse-engineering the business from the assumptions
Before you touch a model, ask: what does a 12% revenue CAGR for this business actually require? For a hospital chain, it requires opening new beds. New beds require capex. Capex requires either debt (watch the interest coverage ratio) or equity (watch the dilution). If revenue is growing 12% but capex is flat, the model is lying about something.
Real-world scenario:
Axis Capital was advising on a healthcare services company IPO in 2023. The company's management projections showed 20% revenue growth with minimal new hospital beds — essentially projecting higher utilisation at existing capacity. The model initially accepted this at face value. The analyst who caught it asked: at what utilisation rate are we currently operating? The answer was 87%. A 20% revenue growth on beds with 87% utilisation requires either significant price increases or new beds. Neither was in the projection.
The model was rebuilt with a capacity expansion schedule. The IPO pricing was adjusted accordingly.
The three-statement mechanics that most models get wrong:
-
Working capital assumptions: Most models use a fixed DSO/DIO/DPO assumption applied to revenue/COGS. For businesses with rapid revenue growth, the change in working capital becomes a meaningful cash drain that is frequently understated. Always check: if revenue doubles in three years, does the absolute working capital investment look right?
-
Deferred revenue and customer deposits: SaaS companies and subscription businesses receive cash before recognising revenue. The deferred revenue movement on the balance sheet is often omitted from cash flow statements in template-based models.
-
Minority interest treatment: When a company owns 70% of a subsidiary, its P&L consolidates 100% of the subsidiary's revenue and expenses — but 30% of the subsidiary's net income belongs to minority shareholders. Many models consolidate the P&L correctly but then fail to deduct minority interest from earnings available to common shareholders.
These checks take five minutes. Not running them has cost analysts their credibility in front of managing directors. The model balancing is necessary but not sufficient — operational coherence is what converts a technically correct model into a defensible analytical tool.
DCF Modelling in 2026: What Has Changed and What Remains Wrong
The DCF model is simultaneously the most powerful and the most abused tool in financial modelling. Most analysts learn the formula. Most do not learn where the formula breaks down.
What has changed: the WACC environment
Between 2010 and 2021, most Indian DCF models were built with WACCs in the 12-15% range. By 2024, with risk-free rates having moved significantly and equity risk premiums being actively debated, the same business modelled with a 13% WACC in 2021 might require a 16% WACC in 2024 — which can reduce a DCF valuation by 30-40% with no change to the underlying business.
Analysts who learned DCF in a low-rate environment and are now working in a higher-rate one need to relearn their intuitions about what a "reasonable" WACC looks like for different business types.
The perpetuity growth rate: the assumption that breaks most models
The Gordon Growth Model terminal value formula — FCF × (1+g) / (WACC − g) — is exquisitely sensitive to the spread between WACC and g. When WACC is 12% and g is 4%, the denominator is 8% — stable. When WACC is 12% and g is 10%, the denominator is 2% — the terminal value is five times larger, and 1% change in either assumption moves the valuation by 50%.
Most analysts use a terminal growth rate of 3-5% "because that's the convention." In 2026, defending a terminal growth rate requires actual reasoning — not convention.
The questions every DCF must answer before being presented:
- What GDP or inflation rate is embedded in your terminal growth rate, and is it defensible for this specific business?
- At your modelled WACC, what implied return on incremental invested capital does your terminal value assume? If it is above the company's historical ROIC, the model is implicitly assuming the company improves. That needs justification.
- What portion of total DCF value comes from terminal value? If it is above 75%, your model is primarily a bet on long-run normalisation, not near-term cash flows. Present it that way.
In 2023, a boutique IB in Mumbai was pitching a consumer electronics distribution company to a strategic buyer. The DCF showed a value range of ₹2,200-2,800 Cr. The senior banker reviewing the model asked: "What's the implied ROIC in your terminal year?" The analyst had never calculated it. When they did, the answer was 28% — significantly above the company's 15-year average ROIC of 16%. The terminal value was implicitly assuming that the company would dramatically outperform its historical capability. The DCF was rebuilt with a more conservative assumption. The value range changed to ₹1,600-2,200 Cr — which turned out to be closer to where the deal ultimately priced.
M&A Merger Modelling: The Mechanics That Trip Up Experienced Analysts
Merger models are where the skill requirements escalate significantly. A standalone DCF tests your modelling mechanics. A merger model tests whether you understand deal structure, purchase accounting, and how to communicate accretion/dilution in a way that a deal team can use.
The purchase price allocation: where most models are incomplete
When an acquirer pays more than a target's book value, the excess must be allocated to identifiable intangible assets and goodwill in a process called purchase price allocation (PPA). Most analysts model the purchase price correctly but then ignore the downstream PPA implications:
- Identified intangible assets (customer relationships, technology, brand) get stepped up to fair value and then amortised over their useful lives (typically 5-15 years)
- This amortisation hits the P&L, reducing reported earnings — often substantially in the years following acquisition
In practice, for an Indian FMCG acquisition where the brand is worth ₹500 Cr and is amortised over 10 years, the annual amortisation charge of ₹50 Cr reduces the combined entity's reported PAT by approximately ₹35-37 Cr per year. A merger model that misses this will show higher accretion than the deal will actually deliver.
The synergy modelling problem:
Synergies drive most merger models. They also produce most of the analytical dishonesty in investment banking. The common failure modes:
- Cost synergies are modelled at full run-rate from day one. In reality, integration takes 12-24 months, and synergies ramp in over time.
- Revenue synergies are included without phasing. Industry data consistently shows that revenue synergies are achieved at roughly 50% of the projected rate, while cost synergies are achieved at roughly 75-80%.
- Integration costs are omitted or understated. Restructuring charges, system integration costs, retention bonuses, and advisory fees for post-merger integration typically equal 20-30% of annual synergy savings in the first two years.
When Zomato acquired Blinkit in 2022 for ₹4,447 Cr in an all-stock transaction, merger model analysis had to account for: the different revenue recognition model between food delivery and quick commerce, integration costs associated with merging two distinct operational platforms, synergies from shared marketing and shared restaurant partner relationships, and the dilution from the all-stock consideration. An analyst who only modelled the EPS impact of adding Blinkit's EBITDA to Zomato's would have missed all of this complexity.
Comparable Company and Precedent Transaction Analysis: The Judgment Layer
Running comps and precedent transactions is mechanical. The skill is in the selection, adjustment, and interpretation — and in knowing when the comps are telling you something the DCF is not.
The most common comps mistake in 2026: mixing LTM and forward multiples
Enterprise Value / EBITDA multiples come in two flavours: LTM (last twelve months) and forward (next twelve months projected). Using LTM multiples to value a company on forward financials, or mixing LTM and forward multiples within the same comps set, will produce a distorted range.
For high-growth companies where EBITDA is expected to grow 30%+ over the next year, the LTM EV/EBITDA multiple will be much higher than the forward multiple — because the denominator is smaller. An analyst who averages a set of LTM multiples and applies it to the target's next-year EBITDA has just misapplied the methodology.
The sector premium/discount judgment:
When your comparables span multiple sub-segments, the multiples will reflect different growth profiles and business quality. EV/EBITDA − g (EBITDA growth rate) normalises for growth differences between companies and is significantly more defensible than raw multiples for growth-stage businesses.
Precedent transactions: the control premium question
Precedent transaction multiples are almost always higher than trading multiples for the same business because acquirers pay a control premium — the premium for the right to operate and direct the business. In Indian M&A, control premiums have historically ranged from 20-40% over unaffected trading price.
The 2026 Skill Stack: What Has Been Added to the Non-Negotiables
Beyond the core valuation models, the IB analyst skill stack in 2026 includes capabilities that didn't exist or weren't required five years ago.
Python and automation fluency
Not "data science Python" — but the ability to automate repetitive financial modelling tasks: scraping financial data from BSE/NSE filings, building data cleaning scripts for comps analysis, automating sensitivity tables across multiple scenarios. Analysts who can do this work save their teams hours per week and get staffed on more interesting work as a result.
Scenario modelling for macro risk
In 2026, with ongoing geopolitical volatility, currency risk in cross-border transactions, and interest rate uncertainty, a single-scenario model is professionally inadequate for any transaction with meaningful cross-border exposure. How does a 15% INR depreciation affect a company with 60% USD revenue but INR costs? How does a 200bps rate rise affect the refinancing risk of a leveraged acquisition closing in 18 months?
Data room analysis and rapid financial due diligence
The skill of rapidly identifying the key drivers of a business's financial performance from raw financial statements — without a structured template — distinguishes senior-quality analysts. Key techniques: revenue bridge decomposition, EBITDA normalisation, working capital intensity analysis.
The Model Review Process: What Happens Before a Pitchbook Goes Out
Understanding how senior bankers review models is as important as knowing how to build them.
What a VP or Director is looking for in the first 3 minutes:
- Does the model balance? (Is total assets = total liabilities + equity?)
- Does the cash flow statement properly reconcile the cash balance on the balance sheet?
- Is the revenue growth rate consistent with the capex trajectory and industry context?
- Does the DCF value make intuitive sense given what the comps set shows?
What gets models sent back:
- Hardcoded numbers in formula cells — not labelled as assumptions, not separated into an assumptions section
- Circular references that are not intentional — models that ERROR when you change an assumption
- Sensitivity tables with formulas rather than data tables — fragile to structure changes
- Inconsistent date headers (some columns say CY, others FY, no explanation of the company's fiscal year)
The communication layer:
The model output is only valuable if it is communicated effectively. The structure that works: State the range. Explain which method you weight most and why. Explain what would have to be true for the high end of the range to be appropriate. Explain the key risk to the low end. Then state your recommendation.
This is not a formula — it is judgment applied to numbers. And it is the capability that justifies the analyst's role as AI tools automate more of the mechanical work.
Closing: From Financial Modelling to Deal-Level Thinking
The financial modelling skills covered in this article — three-statement coherence, DCF quality diagnostics, merger model mechanics, LBO transaction-readiness — represent the technical foundation every IB analyst needs in 2026. But technical fluency, as this article has argued throughout, is the floor.
The ceiling is determined by the business judgment that makes those skills useful on an actual transaction. How do you present a valuation range to a client who is anchored on a number that your model does not support? How do you normalise EBITDA in a data room when management is using aggressive adjustments? How do you build a merger model for a deal where the consideration includes an earnout tied to three-year performance?
At Meritshot, the Investment Banking programme is built around exactly this progression: from model mechanics to deal-level application. Students build full three-statement models, DCF analyses, comps, and LBO models on real Indian transactions — Zomato's Blinkit acquisition, the HDFC Bank-HDFC merger, a hypothetical pharma LBO — with the same complexity and the same scrutiny they would face in an actual deal room.
Explore the Meritshot Investment Banking Programme →
This article was written by the Meritshot content team. Meritshot trains professionals in Data Science, AI Engineering, Full Stack Development, Investment Banking, and Cyber Security through hands-on, practitioner-led programmes.
