From business question to decision.

Most research tools ask you to choose a category, pick keywords, or fill out a structured form. Shodh does not.

You give Shodh the business question the way you would describe it to a strategy partner: context, objectives, geography, and relevant background. "Should we enter the European market?" or "Why are our margins declining?" are both valid starting points.

Shodh reads intent, not just keywords, and decomposes the question into a custom set of research domains: specific, directed investigation questions that, taken together, address the problem completely. The domains are decided fresh for every question, not chosen from a fixed list.

The plan also sets a priority and expected source depth for each domain, allocating investigation effort based on how complex the question is. This plan is versioned and saved, then passed on to research.

Once the question is defined, Shodh determines what actually needs to be investigated to answer it, and in what order.

There is no fixed roster of investigation steps. The number and nature of the domains are whatever the plan decides this particular question needs, whether that is three domains or eight.

Each domain in the plan is a specific, falsifiable research question with its own expected sources and depth, so the investigation that follows stays focused on what matters for this decision rather than gathering information for its own sake.

Once the plan exists, Shodh investigates every domain it calls for.

Sources are not generic search. Each domain is routed to sources appropriate to it: business press, regulatory filings, funding databases, customer sentiment, competitive signals, or financial filings, depending on what the domain requires.

Each domain returns findings broken into individual claims, and every claim is traced back to the specific source and snippet it came from. Nothing is asserted without a citation the system can point to.

Each domain is its own checkpoint. If one fails partway through, it retries and degrades gracefully rather than restarting the whole investigation from scratch.

Raw findings are normalized into a structured evidence graph rather than left as free text. Every claim is linked to the source and snippet that supports it, and every source is weighted by reliability and relevance.

After the first pass, dedicated agents cross examine the investigation against its own plan. They look for domains that came back thin, claims that rest on only one source, and claims that directly contradict each other on the same fact, once differences in time period, geography, definition, and methodology are ruled out first.

When a real gap or contradiction is found, Shodh runs a small, targeted follow up investigation aimed at exactly that gap rather than a full rerun. Conflicting claims are surfaced and investigated instead of silently averaged.

An example: the market domain reports strong demand growth. The competitor domain reports that two leading incumbents both lost funding this year. That contradiction gets investigated directly, rather than smoothed over, because it is often where the real insight is.

Shodh compiles everything gathered and verified across every domain into an auditable investigation.

Every claim can be traced back to the exact source, snippet, and reliability tier it came from. Confidence is not a flat label. It reflects how many sources back a claim and how reliable those sources are.

The investigation also carries forward what the verification step found: which gaps got closed through follow up research and which questions remain open because they could not be verified from available sources.

From there, Shodh moves from evidence to a diagnosis of the problem, a recommendation for what to do, and an implementation plan for acting on it.