Diamond Demand Intelligence

Demand is the part of the market nobody records

Almost every system built for the diamond trade describes supply. Stock lists, inventory databases, price feeds and trading platforms all answer the same question: what is available and at what price. That is useful, and it is also the half of the market that is already written down.

Demand is different. It is spoken and typed, not filed. A buyer who needs eight rounds in a specific colour and clarity range for a customer order writes a sentence into a dealer group, gets what they need within the hour, and the sentence scrolls away. Nothing about that request is stored anywhere in a form that a seller can search later. Demand intelligence is the attempt to capture that layer before it disappears.

How it differs from a marketplace or listing site

A marketplace is a place where sellers publish inventory and buyers search it. The seller's job is to be listed, priced competitively, and hope the right buyer runs the right search. The seller learns nothing about the buyers who searched and did not find a fit.

Demand intelligence inverts that. The unit of data is the request, not the stone. A seller does not ask "is my stone visible?" but "who asked for something I already hold, and when?" That difference matters most for the stones that sit longest: goods in unusual sizes, mixed parcels, or specifications that are perfectly saleable but only to a narrow set of buyers who happen to need them this week.

BrilliantBot is not a marketplace, an exchange, a broker, an escrow service or a payment processor. It does not hold funds and does not settle transactions. It identifies who asked for what and which of your stones fit; price, terms and the deal itself stay between the two businesses.

What a demand intelligence system has to do

• Observe a defined source. Requests have to come from somewhere specific and disclosable. For BrilliantBot that means the diamond trading groups the bot has been added to — no more, no less.
• Recognise a request. Group traffic is mostly not demand: it is offers, chatter, images and price talk. A request has to be separated from everything around it.
• Normalise the language. "1.5–1.7 rd F+ VS2 up GIA" and "looking for one and a half to one seventy round, F or better" describe the same thing in two languages and two notations.
• Compare against real stock. Structured demand is only actionable next to a structured, current inventory record.
• Deliver in time to act. A request answered tomorrow is not answered. Notification speed is part of the product, not a detail.

Where the approach has limits

Being explicit about limits is part of being useful. Coverage is bounded by the groups the bot monitors, so no claim of full-market or industry-wide visibility is made or implied. Free-text parsing is imperfect: a vague request stays vague, and no specification is inferred that the buyer did not state, which means some requests are matched loosely and some are missed.

Inventory quality is the other constraint. If a stock list is out of date, matches will point at stones that have already been sold or placed on memo. And a match is a signal of interest — the negotiation, the price and the relationship remain entirely human work.

Read next

• Methodology — the full pipeline, step by step, including failure modes.
• Diamond buyer demand — how demand actually appears in dealer groups.
• Diamond inventory matching — how criteria are compared against held stock.
• Glossary — definitions of the terms used on this page.