Over the past year, Thurro’s NowCast framework was applied selectively, focusing on areas where high-frequency data has a clear and attributable link to underlying outcomes. This note reflects on two such cases—food & beverage inflation and BSE’s operational revenue—to assess what held up when tested against official data, and what we learned in the process.
Reading food prices before the print
Food and beverage inflation is among the most volatile components of India’s consumer price index. Prices are seasonal, geographically fragmented, and prone to sudden shocks. Despite this, Thurro’s food and beverage inflation NowCast tracked official MOSPI prints closely through 2025, including periods of sharp inflation, rapid disinflation, and brief deflationary episodes.
Most notably, the NowCast was available 10–12 days before the official release. Across months, the model proved to be highly precise—capturing both the peak in late 2024 and the steady cooling that followed into mid-2025.
What made this possible was not model complexity, but structure. The NowCast draws on daily price data from over 1,100 mandis, mapped across 45 food commodities—amounting to nearly 1.5 million price observations per month—and weighted exactly in line with the CPI basket. Over time, the NowCast has ingested close to a billion high-frequency price observations across mandis and official sources—placing it in the same class of large-scale price measurement efforts as the MIT Billion Prices Project, but tailored to India’s CPI structure. Noise reduction techniques were applied carefully, with an explicit bias toward retaining directional shifts.
Thurro’s food and beverage inflation NowCast closely matches MOSPI CPI, capturing the late-2024 peak and subsequent cooling in 2025
The key learning
For macro variables like inflation, the challenge is building a system that can ingest millions of noisy, high-frequency data points and still preserve true directional shifts. When that complexity is handled well, the NowCast becomes valuable not for surprise, but for early, reliable confirmation ahead of the official print.
Capturing BSE’s revenue inflection in real time
BSE’s operational revenue underwent a structural shift beginning in the second half of FY2024, driven by a surge in derivatives activity and broader market participation. Thurro’s BSE NowCast captured this transition as it unfolded, closely mirroring reported standalone operational revenue quarter after quarter through Q2 FY2026.
After years of relatively flat growth, revenue moved sharply higher, eventually crossing INR 9,750 crore. The NowCast tracked this move with near-zero lag, including the inflection itself—not just the expansion that followed.
This accuracy stemmed from the model’s anchoring in direct revenue drivers: trading volumes, notional traded values, derivatives premium turnover, clearing activity, listings, and paid-up capital. Each earnings release was used to recalibrate the system, ensuring the model learned continuously rather than drifting.
BSE’s quarterly operational revenue turns sharply higher from FY2024, with Thurro’s NowCast closely tracking reported results
The key learning
What mattered most in the BSE NowCast was not speed, but grounding—anchoring the model in activity variables that directly map to economic outcomes and recalibrating continuously as new information arrived. That discipline allowed real-time signals to remain aligned with reported results, even as market conditions shifted.
What this year reinforced
Taken together, these cases reinforce that NowCasting is viable across both macro variables, such as inflation, and micro variables, such as company revenues—provided the models are grounded in high-frequency activity data and continuously recalibrated. Over time, this discipline has translated into consistency: historically, around 75% of Thurro’s NowCasts have fallen within a ±5% range of reported outcomes, a level of accuracy that practitioners in finance find far more informative than point forecasts. NowCast is therefore not a tool for distant prediction, but a framework for generating reliable, real-time signals across the economy, one that, in 2025, proved effective across very different use cases.
View disclaimer
See the signal before the market does. From tracking critical economic indicators to benchmarking company performance against peers, contact us to discuss how our premium NowCast models can be tailored to your specific tracking needs
Unlock the power of alternative data
Do not just follow the market — stay ahead of it. Thurro helps you transform raw filings and alternative datasets into actionable insights.
Explore Thurro AltData Book a demo
