Between us we spent twenty years answering questions for procurement, strategy and portfolio teams at some of the largest companies in the world. Three of those questions came back more than any other, and none of them could be answered properly with the tools on the market.
It has several, sitting in different registers in different countries, each carrying its own caveats. The one that matters is the one you can actually act on. Finding it meant repeated analyst briefings, rework, and weeks nobody had.
It moves by grade, by plant, by production run. A cost model that treats it as fixed is wrong in exactly the place a price negotiation is decided, and the assumption sits where nobody can see it.
Consumer insight arrives as bundled bars built from panels that do not agree on who they measured. You can read the chart. You cannot tell who is actually in it.
Each is a different market and a different data problem, which is why they are three products rather than three views of one. What they share is the thing we came here to build: the age of data liberation will not run on black boxes, or on proprietary models nobody outside the vendor is allowed to inspect.
Most intelligence products ask you to trust a number. Ours is written down, versioned, and has two names on it.
Rahul Gotadki, pharma and life sciences · Shubhendra Anand, chemicals, energy and consumer · last revised 6 October 2026
Not a category, a record. Where a figure is derived rather than observed, the derivation is shown and its inputs are named.
Directly observed, inferred from a limited sample, or a gap we cannot close. Each tier has a written definition, used identically across all three platforms.
The variables driving a figure are exposed, adjustable and recomputed in front of you. An assumption you cannot see is a conclusion you cannot test.
Where no primary source exists the platform says so and leaves the space empty. We do not model across a hole and present the result as data.
When our view changes, the change is recorded. A revision is a finding, not an embarrassment.
The standard above is not a promise about conduct. It is a property of the architecture, and it only works because of how the platform is put together.
Registers, filings, tariff schedules, disclosures and review corpora are unstructured and inconsistent. Language models read them, resolve entities across sources that spell the same thing four ways, and emit structured records. Every record keeps a pointer back to the document and the position it came from, which is what makes a named source possible at scale rather than as a promise.
Once a record is structured, every figure above it is produced by explicit computation, not by a model's opinion. The same inputs give the same output on every run, and the chain from source to figure can be walked step by step. This is the line most AI-branded products will not draw, and it is why a Phontix figure can be audited rather than merely trusted.
Precursor ratios, erosion curves, channel weightings and entry dates are exposed as inputs you can move. Change one and the model recomputes in front of you, with the effect on every dependent figure shown. The scenario you need is built in the session rather than commissioned as another study.
Everything on screen is reachable through an API, with source references and confidence tiers carried through rather than stripped at the boundary. Exports are built for the spreadsheets and BI tools your team already runs, and every pull is versioned, so a number you cited last quarter can still be reproduced.
Two founders. Each signs the standard for the domain they know, and runs the demos in it.
Co-founder, Research and Strategy
Eleven years in chemicals, materials, energy and power. Built should-cost models across four Fortune 500 accounts that fed sourcing decisions over more than $500 million of spend, and spent the last two years rebuilding a 38-person research function around AI.
Owns: the Price Fundamentals method and the commodity evidence standard.
Shubhendra on LinkedIn
Co-founder, Pharma and Life Sciences Research
Ten years in healthcare and life sciences intelligence, leading a research portfolio and a team of twenty across six verticals. Most of it client-facing, which is where you learn which parts of a method survive being questioned.
Owns: LOE Edge. The exclusivity layers, the patient-flow models and the evidence standard every estimate meets.
Rahul on LinkedInThirty minutes, run against your molecule, your feedstock or your brand set rather than a prepared dataset, by one of the two people above.
A molecule you are weighing, a grade you are negotiating, a category you are losing ground in. Real, current, and yours.
Every figure taken apart live. Sources named, tiers shown, assumptions moved while you watch to see what the answer is sensitive to.
The session's exhibit, with its sources and method attached, whether or not you go further with us.
No sales engineer, no canned dataset. One of us, your case, and the layers open.
