How big a yeast starter do I need?
Big enough that the yeast multiplies from what you have to what the wort needs. Both halves of that come from somewhere else: the pitch-rate calculator gives the target, and the date on the pack gives what survived. 20.8 L of 1.050 ale wants about 193 billion cells. A 100-billion pack that's 90 days old is down to 37% viability — 37 billion. The starter has to make up the rest.
On a stir plate at 1.037, that gap closes in about 1.15 liters by the Troester model — or 3.5 liters by the White model. Same yeast, same target, same flask of wort, three times the volume. That disagreement is not a rounding error, and it's the part most calculators quietly hide by picking one model for you.
How much DME for a yeast starter?
A starter is just weak wort: 101 grams of dry malt extract per liter makes 1.037, which is the gravity everyone converges on. Weigh the DME, top up to the finished volume with water, boil ten minutes, cool, pitch.
grams of extract = points × 2.72715 × liters · 37 × 2.72715 × 1 L = 100.9 g
| Starter | Grams of DME | Cells from a fresh 100 B pack | ||
|---|---|---|---|---|
| 1.030 | 1.037 | 1.040 | ||
| 0.5 L · 0.5 qt | 41 g · 1.4 oz | 50 g · 1.8 oz | 55 g · 1.9 oz | 151 B |
| 1 L · 1.1 qt | 82 g · 2.9 oz | 101 g · 3.6 oz | 109 g · 3.8 oz | 241 B |
| 1.5 L · 1.6 qt | 123 g · 4.3 oz | 151 g · 5.3 oz | 164 g · 5.8 oz | 312 B |
| 2 L · 2.1 qt | 164 g · 5.8 oz | 202 g · 7.1 oz | 218 g · 7.7 oz | 383 B |
| 2.5 L · 2.6 qt | 205 g · 7.2 oz | 252 g · 8.9 oz | 273 g · 9.6 oz | 453 B |
| 3 L · 3.2 qt | 245 g · 8.7 oz | 303 g · 10.7 oz | 327 g · 11.5 oz | 524 B |
| 4 L · 4.2 qt | 327 g · 11.5 oz | 404 g · 14.2 oz | 436 g · 15.4 oz | 665 B |
| 5 L · 5.3 qt | 409 g · 14.4 oz | 505 g · 17.8 oz | 545 g · 19.2 oz | 806 B |
DME treated as pure extract at 2.72715 g per point per liter (Troester). Cell counts are the Troester model on a stir plate, ±15%. Computed by this page's engine.
Why do two starter calculators give different answers?
Because there are two published growth models and they were built on different ideas. Troester's — the one behind BrewUnited's calculator — scales growth with the food: a stirred starter makes about 1.4 billion new cells per gram of extract, so doubling the flask roughly doubles the yield. White's, from Yeast, scales with the yeast: a growth factor applied to the cells already in there, capped at 6× and saturating at 200 billion cells per liter.
Pitch a healthy amount into a small flask and they agree. Pitch a tired pack into a big one and they don't:
| Starter | DME | Troester | White | Disagreement |
|---|---|---|---|---|
| 0.5 L | 50 g | 108 B | 100 B | 8% |
| 1 L | 101 g | 178 B | 125 B | 42% |
| 1.5 L | 151 g | 249 B | 143 B | 73% |
| 2 L | 202 g | 320 B | 159 B | 102% |
| 3 L | 303 g | 461 B | 183 B | 151% |
| 4 L | 404 g | 602 B | 204 B | 195% |
One 100 B pack at 90 days (37 B live), stir plate, 1.037 starter. Computed from both published models by this page's engine.
At half a liter they're 8% apart, which is inside anyone's measurement error. At two liters — the size most people actually make — Troester says 320 billion and White says 159 billion, a 102% gap. One of those numbers says you're done; the other says you're barely half way.
Neither, to the last billion — but the disagreement is informative. Troester's model has no ceiling, so it keeps rewarding a bigger flask long after real starters stop responding; Brewer's Friend warns it "may overestimate results for very large starters". White's saturation limit is the more conservative assumption. Where they diverge, size the starter on the lower number. Where they agree — small starters from a healthy pack — the answer is trustworthy and the argument doesn't matter.
Stir plate, shaking, or just leaving it?
Agitation keeps the yeast in suspension and the wort oxygenated, and it is worth more than flask size. The same 37 billion cells in the same 2 L of starter wort:
| Method | Growth rate | New cells | Ends at |
|---|---|---|---|
| Left standing swirl when you walk past | 0.40 B per gram | 81 B | 118 B |
| Shaken regularly every hour or two, while awake | 0.62 B per gram | 125 B | 162 B |
| Stir plate continuous vortex | 1.40 B per gram | 283 B | 320 B |
Troester's published rates: 0.40 B/g standing, 0.62 B/g shaken, up to 1.4 B/g stirred. 2 L at 1.037, from a 90-day pack.
Standing still, that 2 L starter grows 81 billion cells. On a stir plate it grows 283 billion — 3.5× more from the same DME. To reach the stirred number without a stir plate you'd need roughly 7.1 liters of starter, which is a bucket, not a flask. That's the whole argument for the stir plate, and it's why shaking it whenever you walk past is a genuinely useful middle option.
When one starter isn't enough: stepping up
A lager is where this bites. That same wort as a lager wants 387 billion cells — double the ale — and one 37 billion pack can't get there in a single 2 L flask under either model. The answer is two steps: grow, let it finish, crash and decant, then pitch that into a fresh starter.
- Step 1: 2 L at 1.037 (202 g DME) — 37 B in, 320 B out
- Step 2: 1.2 L at 1.037 (121 g DME) — 320 B in, 388 B out
Stepping up works because the inoculation rate resets each time. Cram all of it into one huge flask instead and Troester's curve flattens — past 3.5 billion cells per gram of extract the model predicts no growth at all, and in the real flask the yeast runs out of oxygen and simply ferments the wort rather than reproducing in it.
Do you actually need a starter?
Often, no — and the calculators that exist to size starters are not the ones who'll tell you. For that 20.8 L ale you need 193 billion cells. That's 19.3 g of dry yeast at a typical 10 billion cells per gram: 1.7 standard 11.5 g sachets. Two sachets cost a few dollars, need no flask, no stir plate, no DME and no 24-hour head start, and dry yeast holds its cell count for years in the fridge instead of losing 63% in 90 days.
The real reasons to make a starter are strain and temperature, not thrift. Most interesting yeast — the Belgian strains, the English ones with real character, anything a lab released last spring — only comes as liquid. And a lager genuinely needs the cell count that 387 billion implies, which no sane number of packs covers. If you're brewing an American pale ale with US-05, skip the flask and buy the second sachet.
A stirred starter is done in 18–36 hours. You have two choices, and they matter more than the last 20 billion cells: pitch the whole thing while it's actively fermenting (fine when the starter is under about 5% of your batch volume), or cold crash it in the fridge for a day, pour off the spent beer and pitch the slurry. Decant anything bigger than that — a liter of oxidized, under-hopped starter beer is a liter of your batch that didn't have to taste like that.