To calculate colony forming units per milliliter, you count the colonies on a plate, then multiply by the inverse of the dilution factor and divide by the volume of sample you plated. The standard formula is CFU/mL = (number of colonies) ÷ (dilution factor × volume plated in mL). If you counted 150 colonies on a plate that received 0.1 mL of a 1:10,000 dilution, for example, the original sample contained about 15,000,000 (1.5 × 10⁷) CFU/mL. The math is simple, but getting a number you can trust depends on how you dilute, how you plate, what range of colonies you accept as countable, and which organisms you are trying to grow.
The Formula Step by Step
The full expression looks like this in plain language: take the colony count from a plate that falls within the countable range, divide it by the dilution factor you used for that plate, and then divide again by the volume of diluted sample you actually placed on the agar. Most protocols plate either 0.1 mL (for spread plates) or 1.0 mL (for pour plates), so the volume term is usually one of those two numbers. The dilution factor is expressed as a fraction. If you performed three consecutive 1:10 dilutions, your dilution factor at the third tube is 10⁻³, or 1/1,000.
Suppose you spread 0.1 mL of a 10⁻⁵ dilution onto a plate and count 83 colonies after incubation. Plug those numbers in: 83 ÷ (10⁻⁵ × 0.1) = 83 ÷ 10⁻⁶ = 8.3 × 10⁷ CFU/mL. That tells you the undiluted sample contained roughly 83 million viable organisms per milliliter capable of forming colonies under those growth conditions. When you plate replicates, average the counts first and then run the formula once, rather than calculating separately and averaging the final numbers.
Why Serial Dilutions Are Necessary
Most microbial samples contain far too many organisms to count directly. A single milliliter of an overnight bacterial culture can hold billions of cells. Spreading undiluted sample on agar would produce a solid lawn of growth rather than individual colonies. Serial dilution is the standard way to bring the number into a countable window. You transfer a known volume of sample into a known volume of sterile diluent, mix thoroughly, and then repeat the process across several tubes. Each step reduces concentration by a fixed ratio, typically tenfold.
The accuracy of the final CFU number depends heavily on how carefully you perform these transfers. Pipetting errors in either the sample aliquot or the diluent volume propagate through every downstream dilution. One modeling study showed that serial dilution methods maintain accuracy within about ±0.1 log₁₀ of the true value even when individual pipetting errors reach up to 10 percent, and that this holds across starting concentrations ranging from 10⁴ to 10¹² CFU, dilution ratios from 2-fold to 100-fold, and a wide range of plate sizes.1PubMed Central. Estimation method for serial dilution experiments That ±0.1 log₁₀ tolerance sounds forgiving, but it means your result could be off by about 25 percent in either direction under routine conditions. Careful technique, especially consistent mixing at each step, shrinks that window.
A common mistake is shaking the dilution tube too gently or too briefly. Clumps of bacteria that survive intact through a transfer will land on the plate and grow into a single colony even though they contain dozens or hundreds of cells. Vortexing for several seconds between each dilution step breaks most clumps apart. Some organisms, particularly those that form chains or biofilms, need more aggressive disruption, like brief sonication or passage through a needle.
The Countable Range and Why It Matters
Not every colony count is equally reliable. If a plate has only a handful of colonies, random variation dominates; a count of 5 could easily have been 2 or 10 by chance alone. If a plate has hundreds upon hundreds of colonies crammed together, they start competing for space and nutrients, and you can no longer tell where one colony ends and another begins. The traditional countable window is 25 to 250 colonies per plate for spread plates, and 30 to 300 for pour plates, though guidelines vary between regulatory frameworks.
The lower bound of that range has a statistical rationale. One analysis proposed that 25 colonies per plate corresponds to a limit of quantification at roughly a 99 percent confidence level when the limit of detection is between 1 and 5 CFU per plate.2European Journal of Parenteral and Pharmaceutical Sciences. Determining limit of quantification in microbiology – Section: A new approach for the LOQ in Microbiology Below that threshold, you can say something grew but you cannot put a reliable number on how much. Above the upper bound, the plate is designated “too numerous to count” and you move to a higher dilution for your reportable number.
In practice, you should plate multiple dilutions so that at least one lands squarely in the countable range. If you are estimating a sample’s density for the first time and have no idea what to expect, plating five or six consecutive tenfold dilutions covers a wide enough span. For routine work where you have a rough sense of the expected concentration, three consecutive dilutions usually suffice.
Spread Plates Versus Pour Plates
The two classical plating approaches give slightly different results, and the choice affects your count. With the spread plate method, you pipette a small volume of diluted sample onto the surface of solidified agar and spread it evenly using a sterile rod or beads. Colonies grow on the surface, where they are easy to see and count. With the pour plate method, you mix the diluted sample into molten agar before it solidifies, so colonies develop both on the surface and embedded within the medium.
A comparison of the two methods for water-quality testing found that the pour plate procedure was neither as accurate nor as precise as the spread plate procedure for enumerating heterotrophic bacteria.3Journal AWWA. Standard plate count: A comparison of pour plate and spread plate methods One reason is that organisms embedded in agar are exposed to heat when the molten medium is poured. Some species tolerate the roughly 45°C agar just fine, while heat-sensitive organisms take a hit. Another issue is that colonies trapped inside the agar tend to be smaller and harder to distinguish from debris or air bubbles.
Modifications to the pour plate method can close the gap. Lowering the agar concentration and sterilizing the agar separately from the nutrient components improved colony recovery for several test organisms, including Staphylococcus aureus, Salmonella Typhimurium, Candida albicans, and Saccharomyces cerevisiae.4PubMed Central. Improvement of the Pour Plate Method by Separate Sterilization of Agar and Other Medium Components and Reduction of the Agar Concentration The lower agar concentration allows nutrients to diffuse more freely, which helps colonies reach visible size faster. If your lab uses pour plates for regulatory work, these refinements are worth adopting.
How Media and Incubation Conditions Change the Number
The CFU value you calculate is not an absolute count of living cells. It is a count of cells able to grow into visible colonies under the specific conditions you provided: the medium, the temperature, the atmosphere, and the incubation time. Change any of those variables and you can get a dramatically different number from the same sample.
Media composition has a direct physical effect on colony size. Media with high gel content restrict nutrient flow and toxin removal, which suppresses colony development. Excessive agar concentration has even been shown to inhibit the growth of certain bacteria outright.5New Microbes and New Infections. Bacterial culture through selective and non-selective conditions: the evolution of culture media in clinical microbiology – Section: Disadvantages of gelling agents use If colonies stay too small to see or count, your CFU value drops without any real change in the living population. On the other end, using a rich medium with a lower gel concentration can inflate counts by rescuing organisms that would otherwise fail to produce visible growth.
Temperature matters too, of course. Most clinical and food-safety protocols specify 35–37°C for mesophilic counts, but environmental samples often harbor psychrophilic organisms that grow better at 20–25°C. Using the wrong temperature means you are selectively counting a subset of the community. Incubation time plays a similar role: slow-growing species need 48 to 72 hours or longer to form countable colonies, and cutting incubation short leaves them invisible. Whenever you report a CFU/mL value, the conditions under which it was obtained are part of the result.
Where Your Count Can Go Wrong
Even with good technique, plate counting carries inherent uncertainties that you should keep in mind when interpreting your result.
- Clumping: If two or more cells land together and form a single colony, your count underestimates the real number. This is why the unit is “colony forming units” rather than “cells.” One CFU might represent one cell or a cluster of ten.
- Spreader colonies: Motile organisms or excess moisture on the plate surface can produce thin, spreading colonies that merge with neighbors. These make the count ambiguous and often require the plate to be discarded.
- Pipetting inconsistency: Small systematic errors in the volume delivered at each dilution step compound across the series. Calibrated pipettes and fresh tips at each step help, but some residual variability is unavoidable.
- Counting bias: Different people counting the same plate routinely arrive at different numbers, especially near the edges of the countable range where tiny or faint colonies sit on the boundary of visibility. Automated image analysis reduces this subjectivity, which is one of its main appeals.
Because of these factors, the convention in microbiology is to report CFU values as order-of-magnitude estimates. A difference between 2.3 × 10⁶ and 2.8 × 10⁶ is within normal plate-to-plate variation and not meaningful. A difference between 10⁶ and 10⁸ is real and worth investigating.
Cells That Plates Cannot Count
One of the biggest blind spots of the CFU method is that it only detects organisms that can grow under the conditions you set up. Many bacteria enter a state where they are alive and metabolically active but do not form colonies on routine agar. These cells are described as viable but non-culturable, and they pose a real challenge in clinical and environmental monitoring. They are characterized by a loss of culturability on standard plates, which means conventional plate counts underestimate the total viable population in a sample.6PubMed Central. The importance of the viable but non-culturable state in human bacterial pathogens
This is not a rare laboratory curiosity. A wide range of human pathogens, including Vibrio cholerae, Legionella pneumophila, Helicobacter pylori, and pathogenic E. coli strains, can enter this state in response to starvation, temperature shifts, disinfection, or other environmental stresses. From a public health standpoint, this means a water sample returning zero CFU is not necessarily sterile. It may simply contain organisms that the plate cannot detect. If you are working in a setting where the distinction between “truly absent” and “present but unculturable” matters, you need a supplementary method.
Using Optical Density as a Quick Estimate
Measuring how cloudy a liquid culture is, usually at a wavelength of 600 nm, gives a fast and non-destructive estimate of cell density. The idea is straightforward: more cells scatter more light, so absorbance goes up. But converting an optical density reading to CFU/mL requires a calibration curve that is specific to the organism, the growth medium, and the instrument you are using.
The relationship between OD₆₀₀ and viable count is not linear across a wide range of cell densities. One study fitting curves for four bacterial species found that a fourth-degree polynomial provided the best fit between OD and viable counts across the full dilution range tested.7PubMed Central. Estimating microbial population data from optical density Using a simple straight-line conversion, which many labs do by habit, works reasonably well in the low-to-mid density range but introduces substantial error at higher densities where light scattering saturates.
Strain-level variability is another complication. A study of five environmental Pseudomonas fluorescens isolates found that calibration curves differed from strain to strain, with logarithmic models producing R² values between 0.88 and 0.97 depending on the isolate.8Turkish Journal of Agriculture – Food Science and Technology. AI-Assisted Plate Counting for Accurate OD600–CFU Calibration in Some Wheat Rhizosphere-Related Pseudomonas fluorescens Isolates If you apply a calibration curve built from one strain to a different strain of the same species, your CFU estimate can be off by a factor of two or more. The lesson is that OD is a useful shortcut for monitoring growth in real time, but the curve linking it to CFU/mL needs to be built fresh for each organism and medium combination.
There is also a conceptual mismatch between OD and CFU that is easy to overlook. OD responds to every particle in the light path, living and dead alike. After antibiotic treatment, for instance, OD might stay high while viable counts plummet because dead cells still scatter light. One research group comparing calibration methods found that CFU-based calibration underestimates total cell numbers by more than an order of magnitude compared to flow-cytometry-calibrated methods, suggesting that a large fraction of cells present in a culture simply never form colonies.9Communications Biology. Robust estimation of bacterial cell count from optical density If your goal is to know “how many viable, colony-forming organisms are in this tube,” OD is an approximation layered on top of an approximation.
Flow Cytometry and Other Rapid Alternatives
When turnaround time is critical, waiting days for colonies to grow is impractical. Flow cytometry, which passes individual cells through a laser and counts them one at a time based on light scatter and fluorescent staining, delivers results in minutes rather than days. A comparison of flow cytometry and traditional plate counts for monitoring dialysis water found that flow cytometry offered higher sensitivity than heterotrophic plate counts, potentially enabling earlier corrective actions when microbial contamination spikes.10PubMed Central. Comparison of flow cytometry and heterotrophic plate count methods for dialysis water microbial monitoring
Flow cytometry counts total cells, not just those capable of forming colonies. Combined with viability dyes that distinguish live from dead cells, it gives a more complete picture of a microbial population than plates alone. The tradeoff is cost: a flow cytometer is a significant capital investment, and the consumables per sample are more expensive than agar plates. For high-volume or high-stakes monitoring programs, the speed and sensitivity justify the expense. For a teaching lab running routine plate counts, flow cytometry is overkill.
ATP bioluminescence assays offer another rapid alternative. They measure adenosine triphosphate, an energy molecule present in all living cells, and convert the signal to a relative light-unit reading that correlates loosely with cell numbers. These assays are common in food-processing sanitation checks because they give a pass/fail hygiene result in seconds. Their limitation for quantitative microbiology is that ATP content varies enormously between species and growth phases, so the correlation with CFU is rough at best.
Automating the Colony Count
Whether you are counting colonies by eye or using a digital method, the number you feed into the CFU formula needs to be accurate. Manual counting is the default in most labs, but it is slow and subjective. Two technicians counting the same crowded plate can disagree by 10 to 20 percent, especially when colonies are small, overlapping, or embedded in agar.
Automated image analysis software addresses this by applying consistent edge-detection algorithms to photographs of plates. One study developing an open-source ImageJ macro for colony counting demonstrated that the automated approach was faster and more accurate than other open-source digital tools and manual counts across a variety of cellular assays.11PubMed Central. High-Throughput Method for Automated Colony and Cell Counting by Digital Image Analysis Based on Edge Detection Free tools like ImageJ and CellProfiler are available to any lab with a decent camera and a computer, making automation accessible without expensive dedicated colony counters.
The catch is that automated tools require tuning. You need to set parameters for minimum colony size, contrast thresholds, and circularity filters so the software distinguishes colonies from plate imperfections, air bubbles, or precipitate in the medium. Once those parameters are dialed in for your particular organism and plate type, the software is both faster and more reproducible than a human with a marker pen. If your plates look different from run to run, though, you may find yourself adjusting settings constantly, which eats into the time savings.
Fungal and Yeast Counts Require Extra Patience
Everything discussed so far applies to bacteria and to fungi in principle, but fungal and yeast CFU determinations come with additional quirks. Yeasts like Saccharomyces cerevisiae and Candida species generally behave well on plates, forming discrete round colonies similar to bacteria, though they tend to grow more slowly and may need 48 to 72 hours of incubation before colonies reach countable size. Filamentous molds are a different story. A single fungal spore can produce a sprawling, fuzzy colony that covers a quarter of the plate, making it easy to count but hard to get enough individual colonies in the countable range. You often need higher dilutions or smaller plating volumes to keep mold counts manageable.
Growth curves for fungi also behave differently from bacterial growth curves when you try to correlate OD with CFU. Filamentous growth does not increase turbidity in a predictable way because hyphae clump, fragment, and settle unevenly. For mold quantification, direct microscopic counts using a hemocytometer or Coulter counter are sometimes more practical than plate counts, even though those methods count total particles rather than viable units. If you are working with filamentous fungi and need a viability-based number, the plate count remains the standard, but expect wider variability between replicates than you would see with bacteria.
When Only a Few Colonies Appear
Low counts create a dilemma. If you plate a sample and get 3 colonies on the lowest dilution plate, you have a detection, but the statistical uncertainty around that number is enormous. The Poisson distribution governs the random scatter of colony counts, and at very low numbers, the 95 percent confidence interval stretches from less than 1 to nearly 9 for an observed count of 3. Reporting a precise CFU/mL from that plate gives a false sense of accuracy.
Regulatory frameworks handle this differently. Some require you to report the count as “estimated” when it falls below the lower bound of the countable range. Others allow you to use the plate and note that the result has reduced confidence. In pharmaceutical microbiology, the distinction between a countable result and one that falls below the limit of quantification has practical consequences for batch release decisions. A plate with 20 colonies might lead to a very different regulatory outcome than one with 30, even though the true microbial load might be identical.
If your application demands reliable quantification at very low concentrations, membrane filtration is often a better approach than direct plating. You pass a large volume of sample through a filter that traps bacteria, then place the filter on agar and incubate. This concentrates organisms from, say, 100 mL of water onto a single surface, effectively boosting your detection limit by two orders of magnitude compared to spreading 0.1 mL of the same sample directly.
Replicates and How Many You Actually Need
Plating in duplicate is the minimum for most protocols, and triplicate is better. The reason is not just about catching pipetting errors; it is about characterizing the inherent randomness of colony distribution. Even when technique is perfect, the number of cells that happen to land on a given plate follows a random distribution. Two plates from the same dilution can easily differ by 15 to 20 percent just from sampling noise. Averaging across replicates smooths that noise and gives you a more defensible number.
If you are reporting results for a regulatory submission or a publication, include the standard deviation or coefficient of variation alongside your mean CFU/mL. A result of 5.2 × 10⁴ ± 0.8 × 10⁴ CFU/mL tells the reader something useful about reproducibility. A result of 5.2 × 10⁴ CFU/mL standing alone could be beautifully reproducible or wildly scattered, and the reader has no way to know. In research settings, plating each dilution in triplicate across at least two independent experiments is the general expectation. For routine quality-control monitoring, duplicate plating at two or three dilutions is usually sufficient.