Here's the short version up front, because I hate articles that make you scroll to the bottom for the payoff: in the wholesale trade we work in, a published number can only ever be an opening position, and the useful thing is not the number but the conversation that produces it. The mistake almost everyone makes with a price change is treating it as an announcement instead of a conversation — and the mistake the other camp makes is concluding from that that prices should be hidden. They shouldn't. A public price page is the right answer far more often than the people selling "contact us for a quote" want to admit.

Both halves of that matter, and most writing on pricing only gets one of them.

Why I think about this at all

A good part of what we do is export: we handle the export side for lab-grown gemstone factories in China, alongside our own-brand storefront. Under that storefront there is a pricing and quoting system we built, sitting beneath the product structure and the SKUs — and building it is what taught me that the thing people call "a price" in this trade is not a number at all. It's a bundle of commitments, and every one of them is negotiable.

If you want to see this stated plainly by someone with no stake in my argument, read the US International Trade Administration's own description of a pro forma invoice. It calls it "a quote in an invoice format", and it says the document should carry: per-unit prices and totals, weights and dimensions, any applicable discounts, the sales terms or Incoterm including the delivery point, the payment terms, the estimated shipping date, and a validity date.

Read that list again. A price without those last five things is not a quote; it's a rumour. The government's own export paperwork template assumes your number expires, assumes it depends on where you hand the goods over, and assumes it moves with how and when you get paid.

The Incoterms point is the one outsiders underestimate. The International Chamber of Commerce says it better than I could, in the official introduction to Incoterms 2020: "The reason for the frequent misuse of the wrong Incoterms rule is that Incoterms rules are frequently regarded exclusively as price indicators: this or that is the EXW, FOB, or DAP price... Incoterms rules are not, however, exclusively, or even primarily, price indicators."

What the three letters actually fix is the point where risk passes from us to the buyer — and that same point decides which cost lines fall on which side. Carriage, export clearance, insurance level, import clearance, duty. Under EXW, risk transfers before the transport cycle even starts, and we have delivered whether or not the goods ever arrive. Under DDP, we have delivered only if they actually arrive, and we owe obligations — import clearance — that can only be performed inside the buyer's country. The ICC warns traders off both extremes for exactly that reason and recommends FCA instead of EXW, DAP or DPU instead of DDP.

So: same stones, same grade, same quantity, four different Incoterms, four genuinely different amounts of money, and the gap between them is not a markup — it's a different allocation of work and risk. You cannot put that on a page. You can only work it out with the person on the other side.

Then add the rest of what a real deal negotiates. Grade. Certification, and *whose* certification. Volume and the break points. Payment terms — and those are a price, not a formality. The ITA lays out the spectrum honestly: cash in advance means we "avoid credit risk because payment is received before the ownership of the goods is transferred", and is simultaneously "the least attractive option for the buyer, because it creates unfavorable cash flow"; open account is "one of the highest risk options for an exporter" and "one of the most advantageous options to the importer"; a letter of credit sits in between and costs the buyer bank fees. Somebody is carrying that risk and somebody is paying for it. Which one, and how much, is the conversation.

There is no list price to publish, and that is a verifiable fact

I want to be careful here, because this is the part where people assume the exporter is being cagey on purpose. We're not. In this trade there is genuinely nothing to publish.

Natural diamonds have the Rapaport price list, which the trade has quoted off for decades. Read Rapaport's own documentation of what that list is, though: it covers round and pear shapes only, 0.01 to 10.99 carats, well-cut white stones, and Rapaport describes it as "a benchmark of sellers' asking prices for cash deals" — asking prices, not transactions. There is no Rapaport list for fancy-coloured diamonds. And there is no Rapaport list for lab-grown.

What the lab-grown trade did instead was quote off the natural list anyway. JCK reported in January 2023 that MV Research found 57% of lab-grown sellers use the Rap list for every sale — a list of asking prices for a different product. Rob Bates's assessment of why was blunt: some prefer the list's "opacity and complexity", "but the most likely answer is habit."

The one recurring benchmark that does exist for lab-grown is Edahn Golan's quarterly wholesale price list, and I'd point anyone serious at it. Its Q3 2026 edition, published 7 October 2026, puts round IGI-certified goods in a band of $73 to $142 per carat, with a 1-carat D VVS at $87 and a G–H VS at $76. It also reports rounds up about 2% on the quarter but down 9% year on year, with 2-carat rounds down 20% against Q3 2025. Golan's own read is appropriately unheroic: the market "may be approaching a floor", the gains are "an early indication of price stabilization, rather than evidence of a fully established upcycle", and "it is too early to declare that the market has definitively reached its bottom."

Now look at what that benchmark is made of, as its publisher discloses: wholesale transactions in India, described as small-volume purchases, no sample counts, four sizes only — 1, 1.5, 2 and 3 carats — and a basis that has changed three times in six years, most recently switching to certified stones in Q1 2026 after dropping everything under a carat in Q2 2024.

Here is my favourite piece of evidence that the trade has no agreed price, and it isn't an anecdote, it's arithmetic. The headline decline since the series began in July 2018 is given as −92% in Golan's own 7 October 2026 commentary, −93% on his list page the same day, and −96% where JCK quoted the same analyst's data on 10 August 2026. Three numbers, one dataset, one quarter. If the people whose job is publishing the benchmark can't agree on the headline to within four points, there is no list price for me to put on a page.

For scale on where this market has been: Golan's own long-run statistics page has mid-2018 1-carat wholesale around $1,800 per carat against roughly $80 in mid-2026 — a decline of more than 97% in seven years. When De Beers announced in May 2025 that it was closing Lightbox and turning its Oregon plant over to industrial production, it cited a roughly 90% wholesale price drop since Lightbox launched in 2018 at $800 per carat. The pressure reaches the neighbouring categories too: Charles & Colvard, the one public company in moissanite, filed Chapter 11 on 2 March 2026 — $19.2m of assets against $10.5m of debt, revenue down from $30m in 2023 to $16m in 2025 — and its executive chairman's affidavit blamed falling lab-grown diamond prices dragging moissanite down with them. Its assets sold at auction for $2.7m in June 2026 and operations ceased.

Against that, consider what "publish your prices" would mean for us.

Certification stopped being comparable last year. On 1 October 2025 GIA replaced 4Cs grading for colourless lab-grown stones with a two-word quality assessment: Premium or Standard. Premium requires VVS or better, colour D, Excellent polish and symmetry, Excellent cut on round brilliants; Standard bottoms out at VS, colour E–J, Very Good polish. Stones that clear none of the Standard minimums get no assessment at all. GIA's president Pritesh Patel put the reasoning on the record: "GIA will no longer use the nomenclature created for natural diamonds to describe what is a manufactured product", and descriptive terms are appropriate because most lab-grown diamonds "fall into a very narrow range of color and clarity." IGI, which by its own reporting certifies something like 65% of lab-grown stones, still grades on the old scale. So a GIA stone and an IGI stone are no longer describable on the same axis. There is no grade-for-grade comparison to publish a price against.

Certification is a material cost, not a rounding error. GIA's assessment is $15 per carat with a $15 minimum, plus $5 if the stone fails to qualify — and their fee schedule effective 1 July 2026 confirms it's still that regime. Set that against Golan's $87 wholesale carat and the certificate is in the neighbourhood of 17% of the stone.

Duty on this stuff moved twice this year, in opposite directions, by two different legal routes — and the printed tariff schedule has not caught up with either. The base rate on cut, unset lab-grown diamonds is Free; everything that actually gets charged is bolted on in Chapter 99, and that is the part that keeps changing. Late last year one of the layers was struck down in court; the schedule a buyer looks up today still carries the invalidated headings. So a buyer who computes their own landed cost from the published book gets a wrong number, and which direction it is wrong in depends on the week. I am not going to print today's total here, because the only number worth having is the one your own broker confirms on the day you ship — which is exactly the point: a landed price is not a fact you can publish, it is an answer someone has to go and get.

And the currency and the freight won't hold still either. USD/CNY sat at about 6.67 on 22 October 2026 against a 52-week high above 7.17 last November — roughly 7% of a dollar-denominated price gone in RMB terms without a single term of the deal changing, which is more movement than the entire 9% year-on-year fall in Golan's round index over the same window. Drewry's World Container Index, the industry's own published number, printed $4,434, $4,351, $4,485 and $4,585 per 40ft container across four consecutive weekly assessments this October. Cut stones travel by secure courier rather than container, so that index isn't our freight line — but it is a fair picture of how stable the shipping component of anybody's landed price is.

Add it up and you get the honest answer to "why won't you just publish a price": because any number I published would be wrong by more than its own margin before the quarter ended, and the person reading it would have no way to know in which direction.

What the conversation actually produces

This is where I think most pricing advice goes wrong, including the advice I agree with. The standard version of "talk to your customers" is aimed at finding a bigger number. That's not what the conversation is for. The conversation's output is a *structure* — and the number falls out of the structure afterwards.

Neil Davidson's *Don't Just Roll The Dice*, which he first published in 2009 and which is still free, is the best short statement of why. He kills cost-plus in two sentences: "You might have spent one hundred dollars developing your product, or a million, but that money is all spent. Gone. It's a sunk cost. What matters now is not how much you've spent, but what people are prepared to pay." And his actual mechanism isn't a number at all, it's segmentation: "If there had been some way to sell the product to each customer at the maximum price that they could afford to pay, you would have been able to sell $1150 of software. That's what versioning is about."

Patrick McKenzie's user survey writeup from 2010 is the clearest demonstration I know of that a conversation yields a structure rather than a price. Around forty people answered; the top complaint was that the software wasn't free; and the finding he singles out is one user asking for a 25-card rather than a 15-card free limit — a sentence about where the fence should sit, which is the whole business model. Two years later, writing up a SaaS pricing change, he opens with "Virtually no one gets pricing right on the first try" and describes replacing per-unit pricing with a handful of usage-segmented tiers and a $99/month floor. The reported result is about 25% fewer signups and roughly double the revenue per visitor, plus two second-order effects he's explicit about: the floor filtered out hobbyists, and the remaining customers were easier to support and more willing to talk about what they needed. That last one is the real dividend. A price that selects for people who will talk to you compounds.

There's a lovely small version of this in an old Hacker News thread about Baen Books, whose readers kept telling them they should charge more for DRM-free ebooks. The response wasn't to raise the price; it was to build a tier where you pay more to read an advance copy before release. A price objection in reverse is product information.

And it cuts the other way too, which is why I don't want to oversell the method. In a 2017 "Ask HN: How do you set prices?" thread, a consultant posting as amorphid describes being afraid to raise his rate, raising the fear with one of his own customers, and being told he was too cheap and should go up 50% — because the customer depended on the service and wanted him to stay in business. He raised 50% and says he doesn't think he lost a single customer. Set against that, babuskov in a 2013 thread documents the opposite with numbers: a home-inventory product selling steadily at $19.99 for three years at about 30 orders a month, raised to $29.99 on general "charge more" advice, then 42 orders across the following three months — about 43% less revenue. His conclusion was that the advice "is just an idea to consider and maybe test, but I wouldn't trust it blindly." Note the difference between those two: one raised a number after a conversation, the other raised a number after reading advice.

Kathy Sierra is worth reading here precisely because she pushes back. In "Listening to users considered harmful?" she quotes a practitioner telling her that listening to members was "maybe 20% of it, but the other 80% was something else", and adds that "people don't necessarily know how to ask for something they've never conceived of". Her resolution is the part I'd keep: don't take what they ask for at face value, "listen not only to what users say, but more importantly to what is motivating what they say." That is exactly the skill in a quoting conversation. A buyer who opens by asking for a lower per-carat number is often telling you something about their payment terms, their customs exposure or their own customer's grade expectations, and the useful move is to find out which.

The quantitative camp pushes back too, honestly. Paddle, which absorbed Patrick Campbell's ProfitWell work, documents its willingness-to-pay method as a Van Westendorp price sensitivity survey run per buyer persona, with explicit warning that "sending out surveys to a random sample of people won't result in quality data." They're allies on "ask, don't guess" and sceptics on "a chat will tell you the number", and they're right to be. Davidson lands in the same place from the other direction: "Despite all the psychology and economics, you ultimately just have to pick a price. Some price — any price — is better than no price."

The blow-ups are never about the number

Look at the public record of price changes that went badly and you find something consistent: they failed on a communication mechanism, not on the amount.

Hetzner raised prices in June 2026 and the thread that followed is the cleanest example I've seen. One customer of "well over a decade", a self-described advocate, leaving: "Not even because of the price increases themselves, but rather because the way in which they've communicated those. They sure knew how to nuke that in record time." The mechanism is in another comment: there was roughly two weeks' notice that pricing *would* change, with no numbers — "People assumed a max of 50% increase, I guess, not 3x." They announced that a price change was coming without the price, so customers filled the gap with a guess and were then shocked by the truth.

Mandrill in 2016 is the same failure in a different shape. The thread is full of people reverse-engineering their own bills off an announcement that didn't tell them: one commenter computing "$420–$550, about a 5x price rise" for his volume, another unable to work out his exposure at all — "If that's the case, then we're looking at more than a +$1000/mo increase!" When your customers have to do forensic accounting to learn what you're charging them, you have not had a conversation.

Patreon in December 2017 shifted payment-processing fees to patrons — 2.9% plus $0.35, charged per pledge rather than batched, announced 13 December for the 18th — and reversed it within days. Jack Conte's letter: "We messed up. We're sorry, and we're not rolling out the fees change." To TechCrunch: "We got in the way of the creators and their fans." The detail that matters for my argument is *when* they found out: after a week of phone calls with creators, who told them patrons were cancelling small pledges to avoid the fee. The conversation happened. It happened second.

1Password in February 2026 managed to turn the conversation itself into the injury. The email, as quoted in the thread: "Please go to my.1password.com/billing to register your approval. If you do not provide consent by your next renewal date on or after March 27, 2026, your subscription will automatically be cancelled at time of your next renewal." Asking for consent and threatening cancellation in the same sentence is not asking. A nine-year subscriber's objection wasn't the money but the feature list attached as justification — "as if I had previously been asked about if I'd want 'AI-powered item naming'."

Now the control case. Backblaze went from $5 to $6 a month in 2019 and then published a postmortem with the numbers, which almost nobody does. Announced 12 February, effective 12 March. Churn over the six months before versus after: 5.38% to 5.75%. Exits citing billing up around 30%. They built an extension program letting people prepay a year at the old rate, which took about six months of senior engineering time and pulled in over 30,000 accounts and roughly $1.8m. They pre-briefed press, wrote the FAQs and canned support answers in advance in case it went badly. And they had killed an earlier planned announcement on the morning of 22 August 2017, because a competitor exited consumer backup that same day and they decided it would look opportunistic.

The outcome of all that preparation was *boring*. Churn moved by a third of a percentage point and some customers just wrote back "finally". That's what success looks like here, and it's why nobody writes case studies about it.

Two honest complications, because the record contains both. A paying Backblaze customer in that same thread thought the whole apparatus was condescending: "You're putting the prices up by a dollar a month... I'm not moving for an extra $12," and he'd have preferred a one-line "sorry guys, we've not raised them in a decade." Backblaze's own Yev conceded the tone "did come off as a bit apologetic". And a consultant posting as edw519 describes the opposite method working for years: "I raised my rates 10% every 90 days. No discussion, no debate, no permission, just, 'Starting October 1, my new rate is x.' If they said no, I'd walk away. That *never* happened." A conversation is a tool, not a guarantee, and it can be overstaged.

What survives all of it is this: in every failure above, the customer could not find out what they were going to pay, or found out too late to have any input. That is the defect. The amount is almost never the defect.

Where a public price page is simply the right answer

I want to give this its strongest form, because the trade I'm in biases me and I'd rather argue against myself than pretend.

Buyers say they want published prices, and they say it consistently. TrustRadius's 2022 buying study, with about 2,185 respondents, found 71% named publishing pricing on the website as the top thing that would make them more likely to buy, and 57% put "no pricing available on website" among the top three things that make them less likely to. Gartner's March 2026 sales survey — 646 B2B buyers — reports 67% prefer a rep-free experience, up from 61% the year before. 6sense's buyer research puts buyers most of the way through their process before they contact a seller, and says buyers make first contact the overwhelming majority of the time. Those numbers are not about pricing theory. They're about what wastes a buyer's afternoon.

PostHog wrote the advocacy case better than I would. On "talk to sales": "Two words that embody everything we hate about modern SaaS products." On the cost: "It's all a colossal waste of time and energy that's better spent focusing on what matters: building better products." And the point I think is actually the strongest one: "No one should ever feel like they're paying over the odds because they didn't haggle hard enough." That is the real indictment. Price discrimination you can't see is a tax on people who are polite.

37signals has run the clean version of this for twenty years. Full prices on the page, no contact-sales gate anywhere — Basecamp from a free tier up through $199 a month billed annually, HEY at $99 a year or $12 per user per month. DHH: "We have zero people in sales, zero key account managers." Their grandfathering policy is in their terms of service, not a blog post: "Sometimes we change the pricing structure for our products. When we do that, we tend to exempt existing customers from those changes. However, we may choose to change the prices for existing customers. If we do so, we will give at least 30 days notice." Jason Fried said the same thing back in 2008 — give ample warning, 90 days say, consider grandfathering longer-term customers — and in 2017, describing the move to a flat $99: "We didn't change our prices on our existing customers, which was the key."

That policy is itself a conversation, just a pre-recorded one. It answers the question a customer would otherwise have to ask, before they ask it, in a document they can rely on. McKenzie makes the mechanism explicit: grandfather existing customers indefinitely, because next month's revenue is dominated by people who already pay you while long-run revenue is dominated by people who don't yet, so the cost of grandfathering is a marketing expense and the benefit is that a price rise becomes a non-event.

So here's my actual position, and it's narrower than either camp's.

Publish the price whenever the price is a fact. If what you sell is the same thing at the same terms to everyone — a seat, a plan, a box, a one-time licence — then the number exists independently of who's asking, and hiding it isn't segmentation, it's friction you're charging the buyer for. "Call us for a quote" on a product that has a list price is a dark pattern in everything but name: it converts the buyer's time into your negotiating advantage. Someone selling access up to $6,000 a month publishes it on the page and says plainly that he did it because he was tired of answering the same questions over and over for enterprise leads. If he can, most people can.

Quote when the price is genuinely a function of the deal. Not "enterprise" as a euphemism for "we'll see how rich you look" — genuinely a function, as in: the number cannot be computed without facts only the buyer has. In our trade those facts are grade tolerance across a parcel, which lab's paper they need, volume and break points, Incoterm and therefore who clears customs and who carries the duty, and payment terms and therefore who finances the shipment. That isn't a pricing strategy, it's arithmetic with missing inputs.

And there's an honest middle the transparency camp sometimes skips. In that same Hacker News thread about published enterprise pricing, a commenter made the procurement point: above a certain size, buyers *want* a conversation and often need a named account manager's email address just to enter you into their system. Published prices can coexist with that. Publish what's computable, name what isn't, and never make someone book a call to learn a fact.

The part that surprised me: most of the conversation happens on the page

Building that storefront, the thing that turned out to matter most wasn't the catalogue. It was the writing around it — the terms, plain-language explainers of what a lab-grown stone actually is, and above all what a certificate does and doesn't prove. Buyers arrive with reasonable questions and a lot of noise in their heads, and the pages that answer honestly do more selling than any product photo.

That looks like a content exercise and it isn't. It's the same conversation, written down once instead of held a hundred times, and the pricing argument of this essay applies to it directly. When GIA stopped issuing colour and clarity grades for colourless lab-grown stones in October 2025 and went to Premium and Standard, every buyer who had learned to read a certificate over the previous five years was holding a mental model that had just expired. Explaining that honestly — including that it means a GIA stone and an IGI stone can no longer be lined up grade against grade — costs us the easy comparison a buyer wanted. It also means the quote we send afterwards is about the right things.

That is the version of pricing-as-conversation that scales. A call doesn't scale and shouldn't; a page that answers the question the call would have started with does.

It's also consistent with how we'd rather work anyway, and I'd rather say that out loud than pretend the preference is purely analytical. No ads, no sponsored posts, no affiliate links, no third-party trackers, and email as the only way to reach me. Those aren't a growth strategy — they're a preference for low-volume, high-trust contact over reach, which is the same preference that makes me think a price should be the output of a conversation rather than a billboard. If you're optimising for the largest number of transactions with the least contact, nothing in this essay is for you. If you expect to be dealing with the same people in five years, the conversation is the asset and the price is a by-product of it.

After thirty years of this, the thing I'd point at isn't a technique. It's that the two failure modes are mirror images. One camp announces a number to people it hasn't spoken to. The other hides a number it could perfectly well publish, and makes people ask for it as a condition of being allowed to buy. Both are the same mistake: treating the price as the message rather than as what's left over after you've understood each other.

What I'd actually do

  1. Write down which of your prices are facts and which are functions. Publish every fact. If a price doesn't move with anything the buyer tells you, you have no business making them ask.
  2. For the functions, write down the inputs — grade, certification, volume, Incoterm, payment terms, lead time — and publish *the list of inputs* even though you can't publish the number. "Here is what your quote depends on" is a public page, and it does most of the work the number would have done.
  3. Put the expiry on the quote. The ITA's own template has a validity date for a reason. In a market where the wholesale index moved 9% in a year and the duty stack changed twice, a quote without an expiry is a promise you'll regret.
  4. Never announce that a price will change without saying what it will change to. Hetzner's customers assumed 50% and got 3x. The gap you leave gets filled with a guess, and the guess becomes the grievance.
  5. Have the conversation before, not after. Patreon ran the right conversation a week late and it cost them a reversal and a public apology. The calls were going to happen either way; the only variable was whether they happened while the decision was still open.
  6. Grandfather by default, and put it in the terms, not in a blog post. 37signals' policy is three sentences in their ToS and it pre-answers the only question a worried customer actually has.
  7. Give long notice, then extend it. Six months of warning and an automatic extension past the deadline costs you a little revenue and buys you the thing that matters, which is not being the company people tell stories about.
  8. When someone argues about the price, find out what they're actually arguing about. Sierra's point — listen to what motivates what they say — is the entire skill. In wholesale the per-carat objection is very often a terms objection, a customs objection or a certification objection wearing a disguise.
  9. Expect the good outcome to be boring. Backblaze's churn moved about a third of a point and some customers said "finally". Nobody will write about it. That's the goal.