In a wide-ranging conversation with MarTechAI’s Brij Pahwa, Kamaljit Anand, Chief Data Scientist and Managing Partner at KIE Square Consulting (recently rebranded with its AI-focused venture KiEVerse.ai ), spoke about the state of data-driven marketing in India, the challenges of direct-to-consumer (D2C) models, the struggle for real-time identity resolution, and why Agentic AI will be central to the next phase of martech evolution.
Over the last decade, Indian marketers have invested heavily in data platforms and AI, yet surveys show fewer than 30 percent describe themselves as truly data-driven. Why does this gap persist, and what role can martech play in closing it?
If you look at telco and banking, they have had the luxury of customer data at a granular level for a long time. They know how customers spend, how much they talk, what they do, and have used this information for both marketing and risk modeling. They were early investors in CRM and built strong mechanisms to leverage data. What they lack is market intelligence, what the customer is doing when not on their platform. That gap is partly being filled by third-party data and AI models, but it is not perfect. For CPG, the issue is even bigger because there is no direct customer-level data. They depend on third-party agencies, retailers, and e-commerce platforms. Only if a brand runs a D2C channel can it truly see customer behavior and build scientific loyalty programs. But D2C has not scaled enough for most players, 80 to 90 percent of their budgets still go to marketplaces, so D2C lacks critical mass. India has 10 to 15 major e-commerce platforms and multiple quick commerce apps, which means brands have to stitch together data from many sources. Bringing this data together is difficult, formats are not homogeneous, and if the data arrives with latency, its usefulness drops. This is where martech and AI can make a real difference.
Are there examples of CPG companies successfully running D2C to get the data they want?
There are examples, like Saffola, that have invested in D2C setups. But it is not a volume business, whenever D2C grows, it competes with traditional channels. As a corporation you have to decide whether to compete with your own distribution partners. The big benefit of D2C is customer understanding. You get first-hand behavioral responses to campaigns and can build much richer insights. But at scale, D2C continues to be a small part of the mix, especially when 80 to 90 percent of budgets go to marketplaces. In the US, retailers like Target, Staples, and Walmart have been successful in pushing private labels online, effectively becoming D2C brands themselves. But they have also disrupted their own traditional models in the process.
Almost every CEO today expects ROI from martech and AI investments. Yet, why do so many companies stop at pilots?
The biggest issue is fragmentation. Large enterprises have too many systems and too many agencies. If your ROI is measured by the same people who are spending your budgets, that is not ideal. You need centralized measurement, either in-house or through independent partners. Spends have shifted significantly in the last decade, 30 to 40 percent of budgets now go to digital, with the split between off-platform (Meta, Google, YouTube) and on-platform (Amazon, Flipkart, BigBasket). Off-platform drives traffic, on-platform drives conversions. Most of this digital spend is now performance marketing, putting pressure on traditional media. Large companies maintain at least 50 percent reach and frequency spend for brand safety, but younger digital-first brands spend almost everything on conversions. The risk is that when you turn off the oxygen of paid ads, your share vanishes. Market intelligence is another differentiator. The best-performing companies are tracking availability, competition prices, discounts, and keyword trends daily. Those that invest in this intelligence get better ROAS and better campaign efficiency.
CDPs in India still struggle with real-time identity resolution across fragmented datasets. What is the right approach, deterministic, probabilistic, or AI-driven graph models?
A CDP is essential if you want to know the true value of a customer. Without aggregating data across physical stores, e-commerce, and quick commerce, you cannot estimate share of wallet. You may think a customer spending ₹8,000 a quarter is top tier, but if their actual wallet size is ₹1.6 lakh, you have only tapped 5 percent of potential. The goal is to find surrogates of share of wallet and rationalize spending. Tata is a good example. With Tata Neu, they are bringing data from BigBasket, Croma, Taj, and other businesses together. They have built the data foundation, but the real opportunity is to plug this into CRM and marketing engines so money is not wasted on the wrong customers.
You recently launched new marketing intelligence products. Can you talk about them?
We built three major solutions. The first is Brandverse, which gives a complete picture of brand equity, covering listings, pricing, promotions both organic and paid, and customer sentiment from reviews. It goes beyond NPS to identify dimension-level gaps like when customers say a moisturizer’s effect does not last long enough and helps brands act on them. The second is MarketVerse, which aggregates performance across multiple e-commerce and quick commerce platforms, showing availability, price promotions, and share of voice in one dashboard. The third is SpendVerse, which helps marketers decide how to split budgets between off-platform and on-platform, across marketplaces, brands, and even down to the keyword level.
How are these models trained?
The frameworks were co-created with CMOs who helped shape what to measure. ML is the backbone. We generate volume and price signals daily to help brands take pricing and promotion decisions. Unstructured data like reviews is read by AI engines. We use LLMs, custom-built SLMs, and MCP model context protocol to contextualize for e-commerce versus quick commerce, where price elasticity behaves differently. Human oversight is still critical, ML drives most of the business impact, with AI improving automation and efficiency.
Agentic AI Is the Marketer’s New Power Tool for 2025
In this part of the conversation, Anand was clear that the future of personalization lies in smaller, task-specific models and agentic systems that act on data. “Campaigns need to be hyper-personalized and even hyper-local,” he said. CPG brands, he noted, have started creating more than 1,200 ad variants from a single campaign to test which ones work best, then allocate spend accordingly. Anand believes LLMs are great when using public data, but when dealing with proprietary enterprise data, as in banking, telco, or government, companies should build contextual SLMs and agentic AI systems that can execute actions like content generation, campaign optimization, and A/B testing in real time.
If you had to design the ideal AI-driven martech stack for the next five years, what would it look like?
Integration is the most important piece. Data ingestion has become a data engineering problem, not just a BI problem. Tools like Copilot with Power BI and Gemini with Looker are strong, but we have found ThoughtSpot and Metabase to be more pliable and easier to integrate with any AI engine. What is missing is a true marketing data lake. Enterprises have data warehouses and enterprise data lakes, but not marketing-specific ones that know what CMOs, digital heads, and e-commerce heads need. A marketing data lake must ingest structured and unstructured data, transform it intelligently, and serve insights on demand. Transformation today is still rules-driven. AI-driven transformation is nascent. The ingestion layer is sorted, but front-end tools have limitations. Building this middle layer is the next big opportunity.
Related Articles
