A brand manager commissions consumer research, waits three months for an agency to deliver a report, and by the time they present it to stakeholders, the trend it identifies is already dead. Consumer behaviour that took years to shift now flips in weeks.
Bengaluru boy Abhilash Madabhushi is tackling this problem head-on.
This October, his company, Consuma, raised ₹12 crore in seed funding to replace that model. It counts Britannia, Godrej, Pepsi, IPL, Rapido and WPP Media among other big brands as its clients.
It delivers research in 30 minutes instead of the industry-standard three to six months.
Traditional research methods fail because the mechanics are flawed. Surveys attract the wrong crowd — people motivated by gift cards, rather than genuine opinions, and their responses tend toward convenience or fabrication.
Even the best focus groups top out at a few dozen well-screened participants, and scaling them to a few thousand doesn’t solve the extrapolation problem, primarily when representing a market of millions or billions of people.
“A trend cycle itself is 60 to 90 days,” said Madabhushi in an interaction with AIM. However, traditional market intelligence reports can take up to six months to compile when agencies are in operation.
Madabhushi says that consumer needs don’t form over months anymore — they spark and vanish within a few seconds of exposure on Instagram or YouTube.
For instance, watching one reel about a specialised pillow instantly made him question how he sleeps.
“Yesterday morning, I didn’t even know the category existed; 10 minutes later, I’d bought it,” he said.
When discovery, evaluation, and purchase all happen inside the same burst of social media exposure, slow research becomes pointless.
The Solution
Madabhushi’s answer to this mess is a system that treats consumer intelligence like a real-time data problem, not a survey problem.
Consuma scrapes through billions of digital interactions, runs them through autonomous analysis models, and produces insight-ready reports in minutes.
Madabhushi argued that the old advantage of traditional agencies, decades of archived datasets no longer matters. What counts today is real-time relevance. Because Consuma’s cost structure is so low, it can produce more fresh reports in a year. “We give you 1,000x more usage of data in one day,” he said.
“We see ourselves as a Palantir or a CIA for consumer intentions.”
The team realised early that a PDF report, an immovable, two-dimensional artefact, was fundamentally at odds with how modern decisions get made.
“Reports are two-dimensional. You can’t interact,” Madabhushi says. They were built that way because agencies needed to control variables.
The team flipped the UX entirely. Every “slide” became a dashboard page you could manipulate: drag data, inspect sources, open the raw conversations behind an insight, even ask a voice agent to talk you through the findings.
Underneath that interface sits the Rapid Research Platform—a multi-agent system designed to mimic, and then exceed, the structure of a human research team. Traditional firms put 10 analysts on a project; Consuma spins up “10 agents that are infinitely duplicatable.”
Each agent specialises the way a senior researcher would, but scale turns them into a workforce that can analyse millions of conversations instead of a few hundred survey responses.
For example, the tool revealed that people who enjoy deep-roast coffee also tend to read romantic novels, an insight that no human team would have stumbled upon because no one can sift through that breadth of behavioural data.
Consuma’s system begins with real-time “smart scrapers” that adapt to each page and collect only conversations relevant to the brief.
Because every query is contextualised, the engine knows where to look and filters out noise before enriching the data to fill gaps — enabling concrete behavioural answers without surveys.
The collected data then moves through a multi-agent pipeline that works like a digital research team. Different agents handle context, analysis and verification, each operating within controlled windows to prevent drift.
A supervisory agent monitors confidence and triggers deeper checks when needed.
‘Deep Research Sucks at Consumer Research’
But there also lies another crucial question that Madabhushi says nearly every VC has asked him: what happens when deep-research tools catch up?
“Deep research relies on one data source only, which is SERP [pages ranked in search engine results]… consumer-driven data is not publicly indexed. Roughly 5% of public social media is indexed by Google,” he said.
This includes the conversations buried in YouTube comments, e-commerce reviews and forum threads — most of which never make it into Google’s index.
“They’re [deep research tools] excellent at secondary market research,” he said, but quite poor at consumer research.
However, Madabhushi believes deep-research tools can’t evolve into what Consuma is building because their core business model doesn’t push them toward the hardest problem: crawling.
This is where Consuma’s technical differentiation sits.
The company has spent years building scraping systems that can reliably pull real consumer conversations from places search engines don’t index — comments, reviews, forums, Reddit threads, and YouTube discussions, while still operating within legal and ethical limits.
“So, how does an e-commerce platform charge one for damages? If I conduct a DDoS-level attack and put undue stress on someone’s servers, it costs them a large amount of budget. I’m not doing it,” he said. Neither is the company training its models on the data.
Another aspect Madabhushi stressed is transparency. Every insight in a report links back to its source, “which is good for the end customer because they understand where the insights are coming from.”
He also clarified that the system never circumvents paywalled content, adhering strictly to material that is already publicly available.
Madabhushi said that reviews on major e-commerce and content platforms are intentionally public. They’re published for anyone to read at no cost, and the platforms themselves benefit from that visibility. As he puts it, this is public terrain — but “if there is a gate, we won’t enter.”
100,000 Employees with PhDs
That said, Consuma still has to navigate the shifting terrain of web scraping. Platforms continually change policies, tighten access, and close gates, as tech giants grow more wary of automated tools, even those operating ethically.
Another challenge is the wave of new tools, APIs and open-source models that make it easier for individual developers to replicate pieces of Consuma’s workflow. As models get cheaper and more capable, barriers to entry inevitably drop.
But Madabhushi doesn’t see that as a threat.
“I have 100,000 employees with PhDs sitting in the US, funded by Sequoia, Lightspeed, Accel, to make my product cheaper,” he jested — referring to the fact that every improvement in foundational models only strengthens the stack Consuma builds on, keeping him several steps ahead.
However, Consuma also sits at a strange moment in the tech cycle, one where building with AI has never looked easier.
“Anyone can just spin up a product on Replit,” Madabhushi said, adding, “What differentiates the product that I’m building on Replit versus you is the intelligence that goes into what I’m building versus what you are.”
The post The Bengaluru Startup Building the CIA for Consumer Intelligence appeared first on Analytics India Magazine.
