Materials, Products & Market Research

Split material and product research into two halves, because Claude's reliability is completely different across them. The first half is structure and reasoning: what criteria should drive this choice, how do these options compare on those criteria, what are the trade-offs, what should I be asking suppliers, what am I forgetting to consider. Here Claude is excellent. It knows the general properties of materials well - that natural stone is durable but heavy and needs sealing, that a particular species of timber moves with humidity, that a fabric's rub count signals upholstery durability - because these are stable, widely-documented facts that appeared everywhere in its training. It reasons about trade-offs fluently and organises comparisons cleanly.

The second half is specific, current, local facts: this product's price today, its lead time from this supplier to your city, its exact fire or slip rating, whether it is still in production, the supplier's contact. Here Claude is unreliable to the point of dangerous, and the reason is structural. Prices change constantly and vary by region, quantity and negotiation; lead times shift with demand; product ranges get discontinued; specs live in a datasheet Claude has probably never seen. A plausibility engine faced with 'what does this cost' does not say 'I don't have current pricing' unless you make it - it generates a plausible-looking number. That number may be a two-year-old figure from training, a wrong-country average, or invented outright, and it wears the same confident face as the material properties it had right.

So the rule is a clean division of labour: Claude for the framework, the source for the figures. Let it build the comparison structure, populate the stable-property columns, and reason about fit - then treat every price, lead time, availability and precise spec as a placeholder to be replaced with a verified figure from a manufacturer datasheet, a live supplier quote, or a current catalogue. This is the same discipline as precedent and code research, applied to the market: ground what you can, and flag the volatile specifics as unverified until confirmed at source. Blur the two halves and you will quote a client a price that is off by forty per cent, or specify a product that was discontinued last year.

The workhorse output here is the material or product comparison matrix - the same instrument as the precedent matrix, tuned for selection. Rows are your options; columns are the decision criteria; and, non-negotiably, there are columns for the volatile facts marked as unverified until you confirm them. The craft is in choosing criteria that actually reflect the decision, so tell Claude the context first: "I am choosing external flooring for a hot, high-footfall entrance courtyard in coastal India - budget mid-range, low maintenance, slip-safe when wet." That single sentence lets Claude propose the right columns - slip rating, water and salt resistance, heat retention, maintenance, cost band, lead time, look - rather than a generic list.

A good prompt keeps the two halves visibly separate:

Compare these 5 flooring options for the context above.
Columns: material, durability/slip, maintenance, look, sustainability notes,
then TWO clearly-marked columns - "indicative cost band (UNVERIFIED)" and
"typical lead time (UNVERIFIED)".

Fill the property columns from general knowledge. For the two UNVERIFIED
columns, give only a rough band and label each as an estimate I must
confirm with a supplier - never a precise figure presented as fact.
End with the 3 questions I should ask each supplier to firm this up.

Two things make this work. First, the UNVERIFIED labelling builds the truth-status into the table itself, so nobody - including future you - mistakes a Claude estimate for a quote. A cost 'band' ('economy / mid / premium') is a far more honest and useful thing for Claude to produce than a spurious precise number, and it is genuinely helpful for early feasibility. Second, the supplier questions turn the table into an action plan: Claude is very good at generating the exact questions that will firm up the soft cells, which is often the real value - it tells you what you still need to find out. From there you fan out to manufacturers and suppliers, get real quotes and datasheets, and replace each UNVERIFIED cell with a sourced figure and a date. The table Claude built in minutes becomes, after your verification, a decision document you can stand behind - and cost and lead-time work belongs to a quantity surveyor or the supplier, not to Claude, whenever the number will bind anyone.

Interior and FF&E research is where this workflow pays off most, because the sheer volume is punishing: a single fit-out can involve hundreds of items - furniture, lighting, fabrics, hardware, sanitaryware, decorative pieces - each needing a look, a spec, a price, a lead time and a supplier, all pulled together into a schedule and a client presentation. Claude can carry a huge amount of that load. Give it a room concept and it will suggest furniture typologies and material directions; give it a pile of product data sheets and it will extract and tabulate the specs into an FF&E schedule; give it your longlist and it will compare pieces on style, materiality, durability and maintenance, and draft the descriptive language for a client board. For turning research into presentable structure, it is a gift.

But interiors sharpen the verification tax rather than easing it, because the volatile facts are the ones a client acts on. Lead times are decisive in fit-outs - a sofa that is beautiful, on-budget and twelve weeks out can wreck a handover - and lead times are precisely what Claude cannot know; you confirm them with the supplier, every time. Prices in FF&E move with trade discounts, quantities and season, so a Claude figure is a feasibility placeholder, never a quote. Availability is the killer: Claude will happily recommend a discontinued light fitting or a fabric that is out of stock for your dye lot, because it has no live catalogue. And provenance and sustainability claims - 'handmade', 'FSC-certified', 'low-VOC' - carry client and ethical weight and must be confirmed against the actual product certification, not Claude's plausible description.

The honest framing for interiors is: Claude gets you to a sourced shortlist far faster, and does the drudge work of specs, schedules and presentation copy beautifully - but every item that goes to a client with a price, a lead time, an availability or a certification claim must be verified at source first. A useful habit is to keep the FF&E schedule with an explicit 'verified?' column and a date against each confirmed figure, so the schedule itself records its own truth-status. That way the speed Claude gives you never quietly becomes a promise to a client built on an invented number - which, in a fit-out running to a fixed handover, is exactly the failure that costs you the job.

Make verification a built-in stage, not an afterthought, and it stays manageable even at FF&E volumes. The pattern is consistent across everything in this module: Claude drafts the structure and the estimates with volatile facts clearly flagged; you fan out to primary sources - manufacturer datasheets, live supplier quotes, current catalogues, the actual certification - and replace each flag with a verified figure and a date; the schedule or matrix carries its own 'verified?' status so the truth-state is always visible. For anything that will bind a client or a contract - a quoted price, a committed lead time, a specified performance rating - the source is the authority and, for cost that matters, a quantity surveyor or the supplier's quote, never Claude.

Claude adds one more genuinely valuable frame to material choice: it is good at pushing you past first cost toward whole-life thinking. Ask it to compare options not just on purchase price but on maintenance, durability, replacement cycle, energy-in-use and end-of-life, and it reasons well about total cost of ownership and embodied impact - the cheap finish that needs recoating every two years versus the dearer one that lasts twenty. This is exactly the kind of structured, multi-variable reasoning it does best, and it lifts a material conversation from 'what does it cost' to 'what does it cost over its life and to the planet' - as long as the specific numbers underneath still get verified. It reasons about the framework; you source the figures that populate it.

Carry the whole module out on one attitude. Across precedents, codes, site and now materials, the pattern never changed: Claude is a fast, well-read, tireless research partner that structures, compares, explains and reasons superbly, and fabricates specific, current, high-stakes facts just often enough - and just as confidently - that you verify everything that matters at the source. Materials and products are the sharpest case because the volatile facts here become prices, lead times and specs that clients and contracts depend on. Use Claude to think broader and faster and to organise the sprawl; use the source, the datasheet, the supplier and the QS to make it true. Structure with Claude; verify at source; and you research a whole palette in the time it used to take to research a corner of one.