What a colony forming unit is, and what it is not
A colony forming unit is one viable propagule that grew into one visible colony on the agar. That is a deliberately careful definition, and every word of it constrains what your number means. A CFU is not a cell: a chain of streptococci, a clump of staphylococci, a fungal hypha fragment and a spore each produce a single colony from many cells. A CFU is not a genome either, which is why a plate count and a qPCR count of the same sample routinely disagree — nucleic acid persists in cells that can no longer divide.
What a plate count does uniquely well is measure the organisms that are alive and able to reproduce under the conditions you provided. That makes it the reference method for water potability, food safety, sterility assurance, disinfectant validation and inoculum standardisation, and it is why the plate count survives despite being slow and labour-intensive. If you need total particles rather than viable ones, count in a chamber with the hemocytometer cell count calculator, or use turbidity with the OD600 to cell density calculator — both count dead organisms as readily as live ones.
Two conventions travel with the number. Results are reported per mL for liquids and per gram for solids, and they are usually reported as a log₁₀ value as well, because a 6-log reduction is a far more natural way to describe a disinfection process than a fall from 1.56 × 10⁷ to 15.6.
How the plate count formula is built
The formula undoes, in order, every step that separated the plate from the original sample. You plated a small volume of a diluted tube; the count on that plate therefore represents a small volume of a diluted sample, and you must multiply back out on both counts.
Undo the plating volume. If you spread 0.1 mL, the plate saw one tenth of a millilitre, so divide by 0.1 — equivalently, multiply by 10. This is where most errors happen: a spread plate usually takes 0.1 mL while a pour plate usually takes 1 mL, and using the wrong one puts the answer out by a factor of ten.
Undo the dilution. If the tube you plated was diluted 1 in 10 000 (that is, 10⁻⁴), multiply by 10 000. Enter the reciprocal N, not the exponent — the calculator wants 10000 for a 10⁻⁴ tube.
Undo the primary suspension, for solids only. A solid sample is weighed into diluent before the series starts. Blending 25 g into 225 mL gives a suspension in which every millilitre represents 25 ÷ 250 = 0.1 g of the original food, so the primary suspension factor is (25 + 225) ÷ 25 = 10. Multiply by that as well, and the result is per gram of food rather than per mL of suspension. This step assumes 1 g of sample occupies 1 mL, an approximation ISO 6887 makes explicitly and one that is close enough for foods near unit density.
Put together, the multiplier is N × D₁ ÷ V, and the viable count is simply the colony count times that multiplier. The multiplier depends only on how you set the assay up, not on what grew — which is why it is worth writing on the bench sheet once and reusing it for every plate in the run.
Worked example: 156 colonies from 0.1 mL of a 10⁻⁴ dilution
You are enumerating a broth culture. You made a ten-fold dilution series, spread 0.1 mL from the 10⁻⁴ tube onto a plate, incubated overnight, and counted 156 colonies.
- Undo the plating volume. 1 ÷ 0.1 mL = 10.
- Undo the dilution. The 10⁻⁴ tube has N = 10 000.
- Primary suspension factor. The sample was already liquid, so D₁ = 1.
- Form the multiplier. 10 000 × 1 ÷ 0.1 = 100 000.
- Multiply. 156 × 100 000 = 1.56 × 10⁷ CFU/mL.
- Take the log. log₁₀(1.56 × 10⁷) = 7 + log₁₀ 1.56 = 7 + 0.1931 = 7.19 log₁₀ CFU/mL.
Check the plate was the right one: 156 sits comfortably inside the 25–250 colony window of FDA BAM Chapter 3, so this plate is the one to report. The 10⁻³ plate from the same series would have carried roughly 1 560 colonies — uncountable — and the 10⁻⁵ plate roughly 16, which is below the window and would have to be reported as an estimate.
Now the same sample as a solid. Suppose it was 25 g of minced chicken blended into 225 mL of buffered peptone water, and you plated 1 mL of the 10⁻³ dilution of that homogenate and counted 178 colonies. The multiplier is 1 000 × 10 ÷ 1 = 10 000, so the result is 178 × 10 000 = 1.78 × 10⁶ CFU/g, or 6.25 log₁₀ CFU/g.
Why the countable range matters more than the arithmetic
The multiplication is trivial; choosing which plate to multiply is the skill. Every published plate count method specifies a countable window, and the reason is statistical on one end and biological on the other.
At the low end, colony counts follow Poisson statistics, so the relative standard error of a count of C colonies is about 1/√C. Twenty-five colonies therefore carries roughly 20% counting error before any pipetting error is added, and four colonies carries 50%. At the high end, crowded plates suppress each other: colonies compete for nutrients, merge, and become impossible to resolve, so a plate that should have shown 600 colonies reads low. Both effects push you toward the middle of the window.
The window itself is method-specific. FDA BAM Chapter 3 uses 25–250 colonies for the aerobic plate count of foods. Standard Methods for the Examination of Water and Wastewater uses 30–300 for the heterotrophic plate count. Membrane filtration methods for coliforms in water commonly use a narrower window still. Select the convention your method specifies rather than the one you remember, because it changes both the range you may report and whether your plate qualifies.
The practical consequence is that you should always plate at least three consecutive dilutions in duplicate. A single dilution is a bet on knowing the answer in advance; three dilutions guarantee that one plate lands in the window. The reference table on this page shows the concentration band each dilution can quantify at your plating volume, which is the fastest way to choose the series before you start.
Colony count, Poisson error, and reportability
| Colonies counted | Relative standard error | Status under a 25–250 window |
|---|---|---|
| 0 | not defined | Report as less than the detection limit |
| 4 | 50% | Estimated count only |
| 10 | 32% | Estimated count only |
| 25 | 20% | Lower limit of the reportable window |
| 100 | 10% | Reportable |
| 156 | 8% | Reportable |
| 250 | 6% | Upper limit of the reportable window |
| 500 | 4% nominal | Crowding makes the true error far larger |
The nominal error keeps falling above 250 colonies, but crowding bias grows faster than counting precision improves, which is why the upper bound exists at all.
Mistakes that put a plate count out by a factor of ten
- Mixing up 0.1 mL and 1 mL. Spread plates usually take 0.1 mL, pour plates 1 mL. This single confusion is the most common ten-fold error in plate counting.
- Entering the exponent instead of the reciprocal. A 10⁻⁶ dilution is N = 1 000 000, not 6 and not 0.000001.
- Forgetting the primary suspension for a solid. Weighing 25 g into 225 mL already dilutes the sample ten-fold before the series starts. Omitting it understates CFU/g by that factor.
- Averaging plates from different dilutions as raw counts. Convert each plate to CFU/mL first, then average. Raw counts from different dilutions are not the same quantity.
- Counting a crowded plate anyway. A plate above the window reads low, not high, because crowded colonies fail to develop. The bias runs in the direction that makes a failing sample look compliant.
- Reporting more significant figures than the count supports. A count of 156 supports two significant figures at best. Report 1.6 × 10⁷, not 15 600 000.
- Ignoring clumping. Vortex or stomach the sample adequately. Chains and clumps each form one colony, so poor dispersion understates the count and is not corrected by any arithmetic here.
Where the plate count sits among the alternatives
Reach for a different method when time, sensitivity or the definition of "viable" is the binding constraint.
Most probable number (MPN) replaces colony counting with replicate tubes scored positive or negative, and reads the density from a statistical table. It handles low densities, turbid or particulate samples and organisms that will not form discrete colonies, at the cost of much wider confidence intervals than a plate count of the same effort.
Membrane filtration concentrates a large volume onto a filter placed on selective agar, which pushes the detection limit far below anything a 1 mL pour plate can reach. It is the standard approach for coliforms in drinking water precisely because the specification is at or near zero.
Flow cytometry with viability dyes and qPCR both return answers in minutes to hours rather than days, but they measure different things: membrane integrity and nucleic acid respectively, not the ability to reproduce. They complement a plate count rather than replacing it, and the discrepancy between them is itself informative about injured or viable-but-non-culturable populations.
Whatever method you use, the dilution arithmetic is the same. Build the series with the serial dilution calculator, and if you are tracking how a population changed between two time points rather than its absolute size, the bacterial generation time calculator and the cell doubling time calculator convert two counts into a growth rate.
Which standard governs your count
For foods in the United States, FDA Bacteriological Analytical Manual Chapter 3 sets the aerobic plate count method, including the 25–250 countable range and the rules for estimated counts. For drinking and waste water, Standard Methods for the Examination of Water and Wastewater sets the heterotrophic plate count and uses 30–300. Internationally, ISO 7218 gives general requirements for microbiological examinations and ISO 6887 governs preparation of the primary suspension and dilutions. All of them use the arithmetic on this page; they differ in the countable window, the diluent, the medium and the incubation conditions.
Key terms
- Colony forming unit (CFU)
- One viable propagule — a single cell, a chain, a clump or a spore — that grows into one visible colony under the stated conditions.
- Primary suspension
- The first dilution of a solid sample, made by homogenising a weighed analytical unit into diluent. Conventionally 1:10, e.g. 25 g into 225 mL.
- Estimated count
- A result reported from a plate outside the countable window, flagged as such because its uncertainty exceeds what the method's validation supports.
- TNTC
- Too numerous to count: a plate so crowded that discrete colonies cannot be resolved. Never report a TNTC plate as a number.
- Detection limit
- The lowest count the assay can report — the value one single colony would represent. Equal to the multiplier used on that plate.
- Log reduction
- The difference in log₁₀ CFU before and after a treatment. A 5-log reduction removes 99.999% of the viable population.