Here is a claim that sounds too simple to be useful, and is in fact one of the most useful ideas in all of buying: every commodity market has exactly one number that tells the truth about value, and the sticker price is almost never it.
That number is the price per unit — the total cost divided by the quantity of the thing you are actually buying. Price per terabyte for storage. Price per gigabyte for memory. Price per milligram for a dosed product. Price per gallon for fuel. Price per watt for solar. Price per kilogram for a raw material. In every one of these markets, the sticker price is a bundle designed to be compared to other sticker prices — which is to say, designed to be uninformative — while the unit price cuts straight to what you are getting for your money.
Learn to see the unit price behind any sticker and you acquire a superpower that works in aisles you have never walked down: you can walk into any commodity market, in any category, and know within one division whether you are being offered a deal or a trap.
This is the lens MarketCrystal applies to everything, and you can see it running live across every commodity board we publish. This piece explains how it works, why it works, and where it applies — which is nearly everywhere.
Why markets hide the unit price
If the unit price is so clarifying, why do markets bury it? Because incomparability is profitable.
When two products can be compared cleanly, they compete on the comparison, and competition on a clear axis drives prices toward cost. When two products cannot be compared cleanly — because they come in different quantities, different configurations, different bundles, under different model numbers — the seller escapes direct comparison and keeps more margin. Confusion is not a bug in commodity pricing. It is a feature, deliberately engineered.
The tactics are consistent across categories:
- Vary the quantity so no two products share a denominator. A 4TB drive next to an 8TB drive. A six-pack of one dose next to a twelve-pack of another. A kit of two memory modules next to a kit of four.
- Bundle multiple attributes into one price so you cannot tell which one you are paying for. A GPU sells memory and compute together. A dosed product sells quantity and concentration together.
- Attach model numbers and brand narratives that imply differences the commodity does not actually have. A gigabyte of DRAM is a gigabyte of DRAM, but the heatspreader and the model name suggest otherwise.
- Emphasize the flattering spec. Marketing highlights whichever number makes the price look justified, never the unit price that would expose it.
Against all of this, the unit price is a single, boring, devastating move. It re-imposes the common denominator the seller worked to remove. It turns a bundle back into a rate. And a rate can be sorted.
The method, in four steps
The discipline is the same in every market. Only the unit changes.
Step one: identify the true unit. Ask what you are actually buying — not the object, but the value inside it. For storage, you are buying capacity: the unit is the terabyte. For a dosed product, you are buying active substance: the unit is the milligram of that substance. For fuel, energy: the gallon (or the joule). For a GPU bought to hold a model, memory: the VRAM gigabyte. The true unit is the thing that, if you had twice as much of it for the same price, would make the purchase twice as good.
Step two: divide. Total price divided by quantity of the true unit. This is the entire arithmetic. Price per terabyte, per milligram, per gallon, per gigabyte. One division.
Step three: filter for the constraints that make units non-interchangeable. Not all units are equal, and this is where naive unit pricing fails. A terabyte on an SMR drive is not the same as a terabyte on a CMR drive for a RAID workload. A milligram of one cannabinoid is not the same as a milligram of another. A gigabyte of DDR4 will not go in a DDR5 slot. So after computing the unit price, apply hard filters: recording technology, active compound, generation, condition, region. These do not replace the unit price; they constrain the set of products whose unit prices are legitimately comparable.
Step four: sort, and read against history. Rank the filtered set by unit price, ascending. The best value is now visible regardless of what the stickers said. Then, if the data exists, place today’s best unit price against its own history. A low number relative to the past is a buy signal; a high number is a wait signal. The unit price tells you which product is the best deal right now. The unit price history tells you whether right now is a good time to buy at all.
That is the whole method. Identify the unit, divide, filter, sort, and read against history. It is almost insultingly simple, which is exactly why it is powerful — simple disciplines get applied consistently, and consistency is what beats a market built to confuse you.
The unit is different in every market — the method never is
The power of this lens is its generality. Walk through a handful of markets and watch the same move work every time.
| Market | The true unit | What the sticker hides | Key filter |
|---|---|---|---|
| Storage (HDD/SSD) | $/terabyte | small-drive penalty, capacity curve | CMR vs SMR, condition |
| Memory (DRAM) | $/gigabyte | kit configuration, speed premium | generation, channel config |
| GPUs | $/VRAM-GB or $/TFLOP | memory-vs-compute bundle | precision, memory type |
| Dosed products | $/milligram active | pack size vs concentration | which active compound |
| Fuel | $/gallon (or $/joule) | grade, additives, location | region, octane/energy content |
| Solar panels | $/watt | panel size vs efficiency | efficiency, degradation rate |
| Raw materials | $/kilogram | purity, form, lot size | grade, purity |
| Bulk food | $/pound (or $/serving) | package size, prep state | grade, form |
Eight markets, eight different units, one identical discipline. The buyer who internalizes this does not need eight buying guides. They need one method and the willingness to ask, in each new market, “what is the true unit here?” The answer is always available, and it is always the thing that determines whether the product does its job.
A worked example across two very different aisles
Take storage and a dosed consumer product — about as different as two markets get — and watch the method collapse both to the same shape.
In storage, you face a 4TB drive at $235 and an 8TB drive at $248. Sticker instinct says the first is cheaper. The unit price says $58.75/TB versus $31.00/TB — the “cheaper” drive costs nearly twice as much for what you are buying. Filter for CMR if you need RAID safety, sort, done. That is precisely what our $/TB hard drive rankings do across live listings, and the same engine drives the GPU $/VRAM-GB board and the DDR5 $/GB board.
In a dosed product market, you face a six-pack at one concentration for one price and a twelve-pack at a different concentration for another. Sticker instinct compares the pack prices. The unit price — dollars per milligram of the active compound — ignores pack size and concentration theater entirely, reducing both to a single rate. Filter for the specific compound you actually want, sort, done.
Different products, different units, identical move. Once you see it, you cannot unsee it, and you will never again compare two stickers when you could compare two rates.
Why this is the future of price transparency
For most of history, computing the unit price across a whole market was too much work to do by hand. You could do it for the two products in front of you, but you could not do it for every product from every seller, continuously, as prices moved. So markets stayed opaque by default — not because anyone hid the math, but because the math was tedious at scale.
That constraint is dissolving. The hard part of unit pricing at scale was never the division; it was extracting the true quantity from messy, inconsistent product listings. One seller writes “4TB,” another writes “4000GB,” a third buries it in a model number. One dosed product says “10mg x 6,” another says “60mg total,” a third makes you read the fine print. Turning that chaos into a clean quantity, across thousands of listings, is exactly the kind of structured-extraction problem that modern language models handle well and that hand-built systems always choked on.
This is why comprehensive, live, unit-normalized price rankings — the thing that was previously done only for a few well-behaved categories like storage — can now be built for any commodity. The method was always sound. What was missing was a way to apply it at scale to markets whose listings are too messy to parse mechanically. That gap is closing, and as it closes, the markets that profited from incomparability lose their oldest advantage.
Mark’s Take: Every buying edge I have ever trusted reduces to the same move: find the unit that determines whether the thing does its job, divide the price by it, filter for the constraints that make units non-interchangeable, and sort. It works on terabytes, milligrams, gallons, gigabytes, watts, and kilograms because it is not really about any of those things — it is about refusing to let a market compare stickers when it could compare rates. Markets spend enormous effort making their products incomparable, because comparability is competition and competition kills margin. The unit price undoes that effort with one division. It is the closest thing to a universal truth-teller that buying has, and the only reason it has not been applied to every market already is that, until recently, the extraction was too hard. It isn’t anymore.
The bottom line
Every commodity market is built around a single honest unit and a sticker price engineered to obscure it. The unit is the terabyte, the gigabyte, the milligram, the gallon, the watt — whatever the thing is that actually delivers the value you are paying for. The sticker is a bundle designed to be compared to other bundles, which is to say, designed to tell you nothing.
The entire discipline of buying well is refusing that bundle and computing the rate underneath it. Identify the true unit. Divide. Filter for the constraints that make units genuinely different. Sort ascending. Read the winner against its own price history to judge timing. That is the method, and it does not care whether you are buying a hard drive, a graphics card, a dosed product, a tank of fuel, or a pallet of raw material.
The market wants you comparing stickers. Compare units instead. It is the one lens that prices everything honestly — and it works in every aisle you will ever stand in.
MarketCrystal provides trend analysis and market commentary for informational purposes only. Nothing in this publication constitutes financial advice or purchasing recommendations. Prices in all markets change constantly; always verify current data before buying. Past pricing does not guarantee future results.
Follow the plumbing.