Pour Rate and Agitation for Pour-Over: How Flow, Pour Height, and Swirling Change the Cup

16 min read · Updated September 13, 2026

How pour rate, pour height, and agitation change a V60 or Kalita brew, what the 2025 avalanche study actually measured, and how to dial it in by taste.

Ask ten people how to pour a V60 and you will get ten numbers, most of them delivered with more confidence than the evidence supports. Pour slower, pour higher, swirl at the end, and never let water touch the paper. Each of these arrives as if it were a single knob with a single correct setting.

Pouring is not one variable. Kettle flow rate, pour height, how wide the stream is, where you aim it, how deep the slurry sits, and how much you disturb the bed afterwards are all separate things that happen to be bundled into the phrase “pour technique.” Changing one usually changes several others by accident, which is why copying a competition recipe gram for gram so often disappoints.

This guide separates them. It covers what a recipe’s timestamps really specify, what a 2025 fluid-dynamics study measured inside a V60, where the press coverage of that study went past the data, and how to use drawdown time as a reading rather than a score.

The short version

• Recipe timestamps do not give you a pour rate. They give you an average that includes any waiting inside the stage. For pulse recipes the two can differ by several times.
• There is no established optimal g/s. The commonly cited 4 to 6 g/s is convention. The recipes on this site span 1.3 to 8.0.
• Pour height changes agitation as much as pour rate does. A falling stream carries more energy into the bed, up to the point where it breaks into droplets.
• Agitation helps until it clogs the bed. Enough to wet and mix everything evenly is good. Past that, small particles migrate and drainage collapses.
• Drawdown time is a diagnostic. A sudden change tells you the bed changed. The absolute number tells you little.

The rest of this article is why, and what to do about it.

Three different numbers get called flow rate

Three separate quantities share the name, and confusing them causes most of the arguments.

Stage-average delivery rate is grams of water added divided by the length of the stage. This is what a recipe’s timestamps encode, and it includes any time you spend waiting.

Kettle pour rate is how fast water leaves the spout while the water is flowing. This is what a flow-rate scale measures, and what the physics of jet impact depends on.

Drawdown rate is how fast brewed coffee leaves the cone. This one is an outcome. Your grind, the fines in the bed, the filter, and the brewer geometry set it, and pouring at 5 g/s does not mean coffee is leaving at 5 g/s. When you pour faster than the bed can drain, water piles up above the coffee instead.

The gap between the first two is easy to see in the recipes already on this site. These are stage-average delivery rates, computed straight from the pour schedules in the recipe timers:

Hoffmann’s one-cup recipe shows the gap directly. His written method specifies each 50 g pour as a 10 to 15 second window followed by a pause, so the kettle runs at roughly 3 to 5 g/s. The timer’s stage windows fold the pauses in, which is why the table reads 2.0 to 2.5 for the same pours. Both numbers are correct. They measure different things.

The 4:6 method is the extreme case. Its stages compute to 1.3 g/s, but nobody pours 60 g of water over 45 seconds. If each pour takes ten to fifteen seconds, the kettle is running at 4 to 6 g/s and idle for the rest of the stage.

For a continuous pour the two numbers nearly converge. Hoffmann’s Ultimate V60 adds 240 g between 0:45 and 1:15, and you are pouring for essentially all of it, so 8 g/s describes both.

This also explains why the familiar 4 to 6 g/s advice is hard to pin down. Jonathan Gagné’s often quoted 5 g/s is tied to his specific kettle, a Stagg EKG 0.6 L, chosen because it produces an 18 to 19 cm stream breakup length, and he gives a working range of 3 to 7 g/s depending on how fast the coffee drains (Coffee ad Astra, 2020). It was never meant as a universal target, and no systematic measurement study has established one.

What the 2025 V60 study found

The most direct look inside a pour-over bed came from Ernest Park, Margot Young, and Arnold Mathijssen at the University of Pennsylvania, published in Physics of Fluids in April 2025.

Coffee is opaque, so they built a model system: transparent silica gel particles, 0.2 to 1 mm, in a 60-degree borosilicate glass funnel roughly the shape of a V60. A green laser sheet lit a slice through the middle and a high-speed camera recorded it at 100 to 200 frames per second.

What they saw was a repeating three-stage cycle they call avalanche dynamics. The jet erodes a crater in the bed. Particles suspend and travel outward, accumulating against the cone wall. The pile then collapses back inward. The cycle repeats for as long as the jet keeps hitting the bed, and it moves material well below the surface rather than just the top layer.

They then ran real coffee: 10 g of grounds, 150 g of water at 95°C, in a clear plastic Hario V60 with Hario paper, poured from a gooseneck kettle at two settings. A thick jet of 3 to 5 mm at about 15 g/s, and a thin jet of 1 to 2 mm at about 5 g/s. Each was poured from 0, 25, and 50 cm above the cone. The TDS values below are read from the paper’s Figure 7, so treat them as approximate.

For the thick jet, height mattered a lot. For the thin jet, every height landed high, which the authors attribute partly to the longer time a 5 g/s pour takes to deliver 150 g. Their recommendation for anyone chasing extraction is to increase “the distance between the pour-over kettle and the cone to maximize mixing” while reducing flow rate, keeping the jet laminar.

One detail from the model-system data complicates the simple reading. The mixing index at the top of the bed rises with pour height, but the erosion measurements show lower pours dig toward the bottom of the cone more effectively. Height buys you surface mixing and costs you depth.

Where the headlines went wrong

The study was covered widely, and the coverage introduced numbers that are not in the paper.

Several outlets reported that the technique lets you use about 10 percent fewer beans for the same strength. Every brew in the paper used 10 g of coffee. A reduced dose was never tested. The figure appears in interviews and press releases, and no experiment in the paper tests it.

Reported optimum heights of “11 inches” or “15 to 20 cm” appear in no experiment either. The real-coffee runs used 0, 25, and 50 cm, and the highest TDS came at the highest height tested, with no optimum identified. The silica gel runs used 2.5 to 22.5 cm, a range that does not overlap the coffee runs at all.

Headlines promising better flavor have the largest gap. The paper measured TDS by oven-drying the brew, and says so plainly: only TDS is reported. There was no tasting panel and no blinding. Writing in Sprudge, Zac Cadwalader made the substantive objection: different compounds extract at different rates, so hitting a TDS number with less coffee means pulling harder on each individual ground, which tends toward bitterness.

The paper is good evidence about a mechanism. It is not a recipe, and the authors do not present it as one. They also flag their own limits: one particle size distribution, a handful of jet diameters, and temperature effects that they speculate about rather than measure.

Pour height and whether the stream holds together

Two pours at the same g/s can agitate the bed very differently depending on how far the water falls. A stream leaving the spout 2 cm above the slurry arrives with almost no added velocity. From 25 cm it arrives considerably faster, carrying more energy into the bed.

The limit is stream integrity. A falling stream of water is unstable and eventually breaks into droplets, and a shower of drops spreads its energy across the surface instead of driving into the bed. Park and colleagues saw this in the model system: at their highest pour height with the thinnest jet, the stream destabilized before impact and the bed was barely disturbed.

The sources agree that height adds agitation while the stream holds together. They disagree about the mechanism, and that changes the advice. Gagné’s model puts the agitation in turbulence and entrained air, so he recommends pouring just below the height where you hear splattering and managing the rest with pour rate. Park and colleagues put it in a laminar jet eroding the bed, and recommend raising the kettle while slowing the pour to keep the jet intact. Barista Hustle’s pour-height tests measured higher TDS and slower drawdown as height went up and noted the result ran against Gagné’s prediction. The public post is a summary; the full data sits in a subscriber white paper.

The mechanism is still being worked out. What every source agrees on is that a g/s figure alone underdetermines what your pour is doing.

A practical version: raise the kettle when you want more agitation, and watch the stream. If it is splattering or arriving as droplets, you have gone past the useful range for that flow rate.

Why more agitation can stall a brew

Agitation helps extraction for straightforward reasons. It wets grounds that would otherwise stay dry, it replaces the saturated liquid sitting against a particle with fresher water, and it evens out concentration differences across the bed. A bed that never gets mixed extracts unevenly, and uneven extraction produces cups that taste thin and sour while still managing to be astringent.

The limit comes from the bed rearranging itself under you. Moving the slurry moves particles, and the smallest particles travel furthest. When they collect in the lower part of the bed or against the paper, they fill the gaps that water was flowing through and the bed’s permeability drops. Drainage slows, sometimes dramatically.

Fines reducing permeability is well established in espresso, where the pressures make it easy to measure (Cameron et al., 2020; Smrke et al., 2024). Modelling work on drip geometry shows that non-uniform flow through a cone produces real variation in local extraction (Moroney et al., 2019). For fines actually migrating in a gravity-fed pour-over bed, the widely cited source is Gagné’s own bench testing, where seven-second spins gave a 5:18 drawdown against 4:28 for two-second spins. That is a single uncontrolled comparison from one bench. The mechanism is plausible and the observation is common, and it has yet to be measured properly in this geometry.

What this means in practice: if increasing agitation makes your drawdown jump, that is the signal to back off, whatever the underlying mechanism turns out to be.

Spirals, pulses, and swirling

Three more levers get tangled up with pour rate.

Where you aim. Circular pouring spreads water and jet energy across the surface, which makes even wetting easier and keeps the bed shape roughly intact. Center pouring concentrates the energy and moves the bed more, which is what the 2025 study used to produce avalanches. Neither has been shown better for flavor. Treat it as a separate dial. And the folk rule about never letting water touch the paper is overstated: Hoffmann’s published Ultimate V60 method tells brewers not to worry about pouring onto the filter.

How many pours. More pulses change several things at once. Total brew time goes up, the bed gets disturbed repeatedly, the slurry level rises and falls, and the water cools between pours. So “four pours extracts more than two” describes one comparison in which several variables moved together, and generalizes badly. Hoffmann’s one-cup recipe uses four staged pours and his Ultimate uses two large ones, and both are meant to make good coffee.

Swirling and stirring. These produce agitation without touching the kettle at all, which makes them useful as an independent control. A bloom swirl does a different job from a late-brew stir. Early on, getting every ground wet is worth a lot. Later, the bed structure has already evolved and the clogging risk is higher. Barista Hustle’s bloom testing compared several bloom agitation methods, including stirring and the Rao Spin, against no agitation and found no measurable difference in extraction. What changed was how much water the bloom absorbed and how consistent the runs were.

V60 and Kalita Wave

The same kettle technique does different things in different brewers.

Kalita’s own description of the Wave is that the flat bed and three holes keep water from accumulating and let it flow at an even rate. That is manufacturer copy with no published data behind it, and timed drain tests across Kalita models range from about 50 to 92 seconds, so the holes are clearly not the only thing setting flow. Body material and how the filter seats matter too.

The practical difference is that a shallow bed is easier to wet evenly, so it needs less help from the kettle. See the brewer comparison for how the geometries differ more broadly, and Sibarist filters for how much the paper alone changes drainage.

Reading drawdown time

Recipes quote a total time, and it is tempting to treat it as the target. Hoffmann’s one-cup method says drawdown should finish around 3:00 and then adds that you should expect some variance, and that taste is the most important thing.

Brew time carries information about the bed. If the same recipe with the same coffee normally drains in 2:50 and today it took 4:15, the bed is passing water more slowly than it did, and the most likely causes are a finer grind or harder agitation.

What brew time does not tell you is whether the coffee is good. A 2:45 drawdown is not inherently better than 3:30. Different coffees, grinders, and papers land in different places, and chasing a number from someone else’s setup mostly teaches you to grind incorrectly for your own.

The useful combination is three things together: what you deliberately changed, what the drawdown did, and how it tastes. Any one of them alone will mislead you.

Grind and agitation are not independent. A coarse bed tolerates a lot of movement before it restricts. A fine, fines-heavy bed can stall under a technique that worked on yesterday’s coffee. This is the main reason competition recipes travel badly. Start from grind size, and read sour and bitter or dry for the taste side.

Run the experiment yourself

The literature does not tell you what your coffee, your grinder, and your kettle do together, and an afternoon of brewing will.

Hold everything constant: one coffee, one grind setting, 15 g to 250 g, same water, same temperature, same brewer and paper. Vary only pour height, in three conditions:

1. Spout close to the slurry, around 2 to 3 cm, pouring at a steady rate you can repeat.
2. Same rate, kettle raised to roughly 10 to 15 cm, stream still intact.
3. Same rate, raised until the stream visibly breaks into droplets before it lands.

Run each condition at least three times. Record total brew time, the time from the end of the last pour to the end of drawdown, final beverage weight, and TDS if you have a refractometer. Then taste them blind, which means having someone else pour them into unmarked cups, because knowing which is which will decide the result for you.

A second round holds height constant and varies rate instead: roughly 3, 5, and 7 g/s. A flow-rate scale makes this easy. Without one, count the seconds for a fixed 50 g addition and accept that you are measuring delivery rate rather than instantaneous pour rate.

Two things to watch for. Drawdown time is the most sensitive readout you have, so record it every time. And three replicates is enough to see a large effect and nowhere near enough to see a small one, so treat a narrow difference as noise unless it repeats.

Track it in BeanBench

Pour variables are exactly the kind of thing that vanishes from memory by the next bag. Brew with a recipe timer, then log the pour height, rate, drawdown time, and tasting notes in BeanBench. After a dozen brews you will be able to see which coffees tolerate a hard pour and which ones stall, which is information no published recipe can give you.

From here, read coffee extraction for the theory underneath all of this, grind size for the lever you should reach for first, or compare the staged approaches in the 4:6 method and Hoffmann’s V60 technique.