True North: The Water We Built for Coffees You Haven't Met Yet

Designed for when you want light and ultralight roasts to taste how they should, rather than to overcome their shortcomings.

---

Magnetic north is not where the North Pole is. The gap between the two is called declination, and depending on where you happen to be standing it can be a single degree or it can be twenty. A navigator who ignores it won’t notice anything at first; the error only shows up later, as the distance between where you meant to land and where you actually did.

Brewing water has the same problem, and almost nobody accounts for it. Water is not a neutral carrier. The minerals dissolved in it decide which compounds come out of the coffee and which stay behind, and then they decide how what came out tastes once it’s in the cup. Every water gives you a reading on a coffee, and most readings are a few degrees off. Unfortunately, if you’ve never brewed the same coffee two different ways, you have no way of knowing which direction yours is skewed.

We built True North to get that error as close to zero as we could manage. Then we cupped it against the alternatives, and the first coffee (a decaf) almost talked us into shipping a different recipe.

Why this was worth building

Golden Fleece exists so that you can find roasters you’d otherwise never hear about, and that premise has an awkward dependency buried in it. When you order a coffee from someone you have never tried, the only thing standing between their work and your judgment of it is your brew water. If the water is editorializing, you aren’t tasting the roaster; you’re tasting the roaster plus an accent you didn’t choose.

That matters more for the coffees we tend to be most excited about. Light and ultralight roasts are where a producer’s actual work is most legible: the varietal, the terroir, the fermentation, all the things that get flattened as the roast level climbs. They’re also the most fragile, for the simple reason that there is more detail available to lose. So we went looking for a water recipe built specifically for that, and found that every obvious candidate was making a choice we didn’t want to make.

A five-minute primer on the minerals

If you already know this part, skip ahead. If you don’t, it’s genuinely worth the five minutes, because everything after this follows from it. Lotus Coffee Products sells a kit of four mineral concentrates that you add drop by drop to distilled or reverse-osmosis water, and they divide into two jobs.

Calcium and magnesium do the extracting. They’re both positively charged ions that grab onto flavor compounds in the coffee and pull them into solution, and together they’re measured as hardness, usually written GH. More hardness means more extraction, and a stronger, more intense cup along with it, right up until the point where it becomes too much.

Sodium and potassium carry the buffer. Both arrive attached to bicarbonate, which neutralizes acid, and together they’re measured as alkalinity, written KH. Alkalinity decides how much of the coffee’s acidity survives to reach you. A little of it smooths off harshness, while too much of it leaves the cup flat and the fruit you paid for simply isn’t there.

The interesting part, and the part most water guidance skips straight past, is that the two members of each pair are not interchangeable. In 2014, Christopher Hendon and Maxwell and Lesley Colonna-Dashwood published a paper in the Journal of Agricultural and Food Chemistry that ran the numbers on exactly this, modeling how strongly each common water ion binds to seven compounds found in roasted coffee (a set of acids, plus caffeine, plus a representative aroma molecule).

Magnesium won every single comparison. It’s a smaller ion carrying the same double positive charge as calcium, so it sits closer to the molecule it’s grabbing and holds on harder. There was also a subtler result tucked into the middle of the paper: magnesium binds the bright acids (lactic, malic, citric, the ones that read as fruit and sweetness) somewhat more preferentially than it binds the harsh ones, compared with how calcium treats the same pair. It’s a modest edge, roughly five percent, but it points in exactly the direction you’d want for a light roast and it compounds across a whole brew.

Sodium was the surprise. It binds coffee compounds so weakly that for caffeine and for eugenol (a warm, woody aroma note) sodium actually manages to be worse than plain water at pulling them out, which is to say it contributes essentially nothing to extraction at all. Potassium is weaker still, which turns out to be a feature rather than a flaw, and we’ll come back to it.

The gap we found

Rather than trust anybody’s summary, we read the actual recipe targets out of the Lotus calculator and lined all of them up next to each other.

All values in ppm as calcium carbonate equivalent, the industry convention.

Every filter recipe Lotus ships is either calcium-dominant or evenly split. Not one of them leads with magnesium, which is the ion the chemistry says does the most work, and does it most selectively in favor of the exact flavors light roasts are prized for. That’s genuinely strange, because magnesium-led water is not some exotic fringe idea. Barista Hustle has built its entire published recipe line on hardness from magnesium sulfate rather than calcium, and has done so for years, so the approach is thoroughly mainstream. It just wasn’t represented here, which is the lane we took.

The Four Decisions

Magnesium leads, at 50 ppm

This is the whole thesis. Because magnesium and calcium are both doubly charged, sixty parts per million of hardness contains the same number of extracting ions either way. We aren’t adding more of them. We’re swapping in ones that grip harder.

That swap has a pleasant second-order consequence, which is that magnesium is the lighter ion (24.3 against calcium’s 40.1) so an equal count of them weighs less in the water. True North carries about 89 mg/L of total dissolved material against roughly 117 for a comparable calcium-led recipe at identical hardness, which is around a quarter less stuff floating in the cup with no loss of extracting power whatsoever. Scott Rao makes the same argument in the notes to his own Lotus recipe, which is that keeping total dissolved solids (TDS) low is what enhances clarity and acidity. Magnesium gets you there for free.

Calcium stays, at 10 ppm

We didn’t go all the way to pure magnesium, because calcium accents sweetness and lends the cup structure and weight, and water built on magnesium alone can read thin, austere, and faintly metallic (technically impressive, and not much fun to drink). Ten parts per million is a seasoning amount that puts a floor under the cup without shifting its character. We’ll be straight with you about the limits of that claim later on, because it’s the one decision our testing didn’t settle.

Sodium is left out entirely

Two independent reasons, and they happen to agree. The first is that sodium does no extraction work (that’s the Hendon finding above, the one where it manages to underperform plain water on some compounds). Adding it means adding dissolved material that raises your TDS reading and contributes nothing in return.

The second is that sodium does have a real effect, just not an extractive one: it suppresses the perception of bitterness and rounds the cup off. That’s a genuinely valuable tool, and on a darker roast or an extraction that has run hot, sodium can rescue a cup. It is also the precise opposite of what we’re after here. Rounding is declination. Every edge it softens is a piece of information about the coffee that doesn’t reach you.

Potassium carries the buffer

You still need some alkalinity, and the only real question is what to attach it to. The answer is whichever ion is the most flavor-invisible, and potassium binds coffee compounds even more weakly than sodium does, which makes it close to an ideal delivery vehicle. It brings the bicarbonate along and otherwise stays out of the way. There’s a nice symmetry to it as well, since coffee is naturally potassium-rich and several of its acids already exist inside the bean as potassium salts.

Fifteen parts per million is low, given that Light and Bright runs 25 and the SCA’s general recommendation is around 40. At 15 the water barely resists the coffee’s acids at all, which is precisely what keeps the acidity legible.

However, we didn’t go to zero, and the reason is the one real tension in this design. Magnesium’s strong binding is indiscriminate, so alongside the acids you want it also drags out more quinic and chlorogenic acid, which is the pungent, astringent end of the spectrum. Light roasts are already the highest in chlorogenic acid of any roast level, because roasting is precisely what degrades it. A magnesium-forward recipe pointed at light roasts is therefore walking straight toward a known hazard, and it needs a counterweight.

Fifteen ppm is that counterweight, and it’s worth holding onto, because it becomes important later in a way we didn’t anticipate.

---

Then we tried to break it

A recipe designed from published chemistry is a hypothesis, not a result. So we built four alternatives, each one differing from True North along exactly one axis, and cupped them blind against it, adding Lotus’s Ultra Light as an outside reference.

A, B and C all sit at the same hardness, so each comparison moves exactly one thing. Every coffee we used came from a roaster on Golden Fleece (which felt like the right way to do it).

The first result said we were wrong

We started with an El Vergel decaf natural from Newbery St., a light-roast Colombian with strawberry and tropical fruit notes.

D won. The highest-hardness cup, the one with the most magnesium, harsh on the first sip and then the punchiest flavor in the lineup by the end. C came in second, and True North placed behind both of them.

C beating A was the worst possible result, because those two differ only in cation ratio: identical hardness, identical buffer, 83% magnesium against 50%. That was the one comparison designed to test the entire premise, and the premise LOST. The obvious move at that point was to raise the hardness, publish the winner, and quietly pretend the whole magnesium argument had never happened.

We asked why instead

Sugarcane EA decaffeination doesn’t only remove caffeine. Published work reports losses of roughly 35–43% across individual chlorogenic acid isomers during EA processing, with one study measuring a 52% drop in total chlorogenic acid. Roasted decaf also runs about 25% lower in trigonelline and 16% lower in carbohydrates than its caffeinated equivalent. Chlorogenic acid, as it happens, is precisely the astringent compound that our 15 ppm of buffer exists to counterweight, and in that particular cup most of it had been stripped out before we ever touched it.

Which explains everything. The guardrail was solving a problem that had largely been removed, so the high-magnesium cup could push extraction hard without paying the usual astringency penalty. Less total soluble material explains why more hardness read as “punchiest” rather than “muddy.” And the extra buffer in cup B had nothing useful left to neutralize, so all it managed to do was mute the fruit.

That gave us a prediction rather than a conclusion: every mechanism that made D win on a decaf is absent in a caffeinated light roast, and if we were right, the ranking should reverse.

It reversed

Next up was an Advanced Natural Gesha from Jhonatan Gasca’s Zarza Bella Vista, roasted by Moonwake. That’s 48 hours of anaerobic fermentation, thermal shock, and a cup the roaster accurately describes as “rich, tannic Concord grape” with acidity that “is high.” A full-chlorogenic, high-acid, tannic coffee, which is to say the exact opposite substrate.

D dropped to mid-pack and turned mineral and spicy as it cooled, which is over-extraction of exactly the tannic fraction the decaf didn’t have. C, which had placed second on the decaf, finished last: “too mineraly,” “earthy,” “doesn’t really have anything going for it.”

True North won.

Both reversals were exactly what the chlorogenic acid mechanism predicted, and the fact that we hadn’t changed the recipe on the strength of that first result turned out to be the most important decision in the whole process.

What buffer actually costs you

The most useful thing we learned wasn’t about our recipe at all. Cup B, the one carrying the extra buffer, produced an identical signature on both coffees and later on a third. Hot, it was the best cup on the table, sweet and floral and delicate and immediately appealing, and then as it cooled it hollowed out into sweetness with nothing behind it. On the Gesha it took four tasting passes for True North to pull ahead of it, and five before it became a clear favorite.

If we had only tasted these hot, we would have shipped the wrong recipe twice.

That’s the practical lesson worth carrying away even if you never make this water: alkalinity trades late complexity for early appeal, and a hot-only tasting cannot see the trade at all. Taste your water experiments as they cool, because it changes the answer.

The sweetness we couldn’t find

True North kept drawing the same criticism, which was that it came out punchy and acidic, with plenty of flavor, and not very sweet. So we ran a third round on a Lumen from Diego Bermudez’s Finca El Paraiso, roasted by Native Coffee Company (80% washed Castillo, guava and cheesecake and pear), a gentler and sweeter coffee than the Gesha. Four new cups, hunting for the missing sweetness.

We tried more buffer at 20 ppm rather than 25, then more calcium, then both of those levers together, and finally we tried swapping part of the potassium out for sodium. Nothing worked cleanly.

Buffer at 20 was just a milder version of buffer at 25. Sweeter, less punchy, “enjoyable but not complex.” Across three coffees the direction NEVER changed. There’s no optimum hiding somewhere between 15 and 25, just a straight line, and every part per million buys you sweetness by spending complexity.

More calcium read mineral, not sweet. We tested it at 20 ppm and at 30, and both times the note was minerality: “off-putting mineraliness,” “too mineraly.” That runs against the conventional framing, and against what we ourselves fully expected going in.

Both levers together were the worst cup of the session. Flat every round, then savory. They simply don’t add up, because buffer strips acid structure while calcium adds mineral weight, and the combination is duller than either one of them alone.

At some point the honest reading is that we were asking water to do a job that belongs to the roast. Ultralight roasting leaves less caramelization behind, and no mineral profile is going to manufacture sweetness that was never developed in the first place. Water redistributes emphasis. It does not add sugar.

So True North ships exactly as designed, not because we couldn’t be bothered to tune it, but because everything we tried to improve it with made it worse on the coffees it’s actually for.

---

The Recipe

Magnesium 50, calcium 10, potassium 15, sodium 0, all in ppm as calcium carbonate, built on top of distilled or reverse-osmosis water.

Before you reach for the droppers, check which ones you actually have. Lotus ships two tip styles and they are genuinely not interchangeable. A round drop delivers about 1.79 times what a straight one does, so the same recipe needs roughly half as many drops. Current kits come with round tips; straight tips still sell as an optional accessory (and shipped with earlier kits), which is why so many older drop counts online assume the straight style. Use the tip photos next to the dropper control in the calculator if you are unsure which you have.

This is worth laboring because nearly every drop count circulating online, including the familiar “13 calcium, 11 potassium” for Light and Bright, is a straight-dropper figure from the original kit. On a round dropper that same recipe is 7 and 6. Round tip, use the first table; straight tip, use the second.

The tables below cover the bottles people actually mix in: the Lotus 450 mL and 900 mL bottles, and a US gallon jug from the supermarket. For any other volume, or to see how close whole-number drops land on the ppm targets, use the recipe calculator on this page.

Round dropper:

Straight dropper:

One caution for round-dropper owners: don’t try to mix this in the 450 mL bottle. A round drop in there is nearly 18 ppm of hardness and the recipe calls for 10 ppm of calcium, which is a target finer than a single drop can possibly resolve. Use the 900 mL bottle or a gallon jug instead, and pour off what you need. On the 900 mL bottle whole-number drops land a little off the exact ppm targets (roughly magnesium 54, calcium 9, potassium 13), which is still far closer than anything 450 mL can do; the gallon is the nearest of the common containers.

The ppm figures are the recipe. The drop counts are just a convenience for getting there.

This is a filter recipe, meaning pour over and immersion. Espresso inverts nearly every requirement (Lotus’s own espresso recipes drop hardness to 20 and push buffer up to 45–55, because at espresso’s concentration and pressure a thin buffer turns harsh and unstable in a hurry), so don’t run this through a machine. Also, because the calcium level here is deliberately low, keep it to kettles and manual brewers generally. Very low-calcium water made from distilled can be corrosive to equipment over time.

The Dial

True North is acidity-forward and not especially sweet. That is the deliberate shape of the thing, not a flaw we failed to fix, and it won’t be everyone’s preference on every coffee. So rather than pretend there’s one right answer, here is the trade, measured.

Add one step of potassium, which takes you from KH 15 to KH 20. In the 450 mL bottle on a straight dropper that is 4 drops instead of 3; the dial on the calculator will make the same bump for whatever volume and tip you are using, and you’ll get a sweeter, rounder, more immediately likeable cup that gives up some punch and some complexity in exchange.

We tested this three times on three coffees and the trade was identical every time. Sweetness up, definition down, in proportion. Where you want to sit on that line is a preference rather than a mistake, and we’d rather hand you the dial and tell you honestly what it does.

If the cup is off

Three levers. Magnesium sets intensity, potassium sets how much acidity survives, and calcium sets the floor (that last one we’d leave alone entirely). Change one step at a time, hold everything else constant, and taste again as the cup cools.

Before you touch the water at all, rule out the brew. Sour and weak together is under-extraction; grind finer. Bitter and drying together is over-extraction; grind coarser. Minerals won’t fix either one.

The targets below are GH and KH in ppm as calcium carbonate, which is what the recipe actually is. Tap one to load it in the calculator, or use the dial there, and it will give you drops for the bottle and tip you have. One step of magnesium is 10 ppm of hardness; one step of potassium is 5 ppm of buffer. In the 450 mL bottle on a straight dropper, that is one drop of each, which is why older guides are written that way.

A note on calcium, since the obvious advice would be wrong. You might reasonably think that a thin or austere cup calls for more calcium (we thought so too, and it is what most guidance implies). We tested it at 20 ppm and at 30 ppm across two coffees, and both times it made things worse rather than sweeter, reading as minerality. Leave calcium at 10 ppm.

Finally, if a cup goes flat on you, check your base water before you change anything else. Residual alkalinity from tired reverse osmosis or an exhausted filter pitcher is the most common invisible confounder in all of this, and it pushes every result in the same direction.

What we’re not claiming

We’d rather be straight with you about the edges of this.

Three coffees is three coffees. It’s more than most recipes get, and all three were blind tastings across multiple passes, but it isn’t a study.

The calcium level is the least tested decision in the recipe. We know that more than 10 ppm makes things worse. We never tested less; nobody ran a cup at zero. The 10 ppm rests on theory and on the general finding that pure-magnesium water reads austere, not on our own data.

Sodium remains genuinely open. We left it out on principle and then, late in the process, tested a cup that swapped a third of the buffer over to sodium. It performed better than we expected: fruity and punchy in every pass, and the only variant that kept its fruit and picked up sweetness. We haven’t run it head to head, so it isn’t in the recipe. But we’re not going to pretend we ruled it out when we didn’t.

And the honest summary of the Hendon paper is that there is no single water composition best for all coffee. There’s water with more extracting power and water with less, and the right answer depends on the bean and the roast it was built for. On top of that, the bicarbonate-selectivity idea we leaned on during the design is one the authors themselves flagged as untested future work.

What we’re confident about is a good deal narrower: if what you want is the least distorted possible reading on a light roast from a roaster you’ve never tried, then the minerals should lead with magnesium, skip sodium, and buffer lightly. That’s supported by the chemistry, it survived every attempt we made to beat it, and it wasn’t previously available as a Lotus filter recipe.

The loose end we’re still pulling

That decaf result never stopped bothering us, in a good way.

The reason True North works is that light roasts are loaded with chlorogenic acid and need a little protection from it. Decaffeination removes a third to a half of that acid before the coffee ever reaches a roaster, which means decaf isn’t simply caffeinated coffee minus the caffeine. It’s a different substrate, and it plausibly wants a different water.

Decaf has a reputation for brewing flat, and we now think a real part of that is people using water tuned for a bean that had more in it. That being said, we only have one session’s data, and we tested a single decaf by a single process. The winning cup was also simply the hardest water we happened to offer. We never went above it, so we don’t know whether we found the peak or just our own ceiling.

That’s not a recipe yet, it’s a promising lead. We’re chasing it down, and when we have something we trust, you’ll find it here as True North (decaf).

The Last Variable

Every bag of specialty coffee is a claim somebody is making about what a coffee can be. A producer made choices, then a roaster made a great many more, and by the time it reaches you a lot of intention has accumulated in a very small amount of plant matter. Water is the last variable in that chain and the only one that is entirely yours, which also makes it the easiest one to get wrong without ever finding out you did.

So mix a batch, brew something you have been curious about, and taste it again once it cools. If it turns out you want it a little sweeter, you already know which drop to add.

---

True North was developed in-house by Golden Fleece Coffee for the Lotus four-mineral kit. The calculator on this page will give you drop counts for your volume and dropper style, and you can pick up the kit from Lotus’s shop on Golden Fleece. The coffees used in testing came from Newbery St., Moonwake and Native Coffee Company.

Sources

• Hendon, C. H., Colonna-Dashwood, L., & Colonna-Dashwood, M. (2014). The Role of Dissolved Cations in Coffee Extraction. Journal of Agricultural and Food Chemistry, 62, 4947–4950. doi.org/10.1021/jf501687c
• Farah, A. et al. (2006). Chlorogenic Acids and Lactones in Regular and Water-Decaffeinated Arabica Coffees. doi.org/10.1021/jf0518305
• Lotus Coffee Products recipe calculator (their published kit recipes; True North lives on this page)
• Lotus Coffee Products on dropper variability
• Barista Hustle, DIY water recipes