Shrimp Postlarval Volumetric Counts and Sampling Bias

The basic problem: the shrimp are moving, but the number is only an estimate

A shrimp hatchery may report that a bag contains two hundred thousand postlarvae, yet the bag was never counted animal by animal. Often the number is inferred from a small portion of water, a concentration step, or repeated comparison with a counted reference. Translating such a record means distinguishing measured subsample counts from estimated total abundance. The calculation is simple; defending the sampling assumptions is the hard part. If those assumptions fail, arithmetic can produce a precise-looking but misleading stocking figure.

Postlarvae, commonly abbreviated PL, are developmental stages after the earlier larval phases. PL numbers are important when a hatchery supplies seed to nurseries or grow-out ponds. The reader wants to know how many animals were actually transferred, how that number was estimated, and how confident the estimate is. A count sheet that omits volume, mixing method, sample position or replicate information cannot fully answer those questions. Saying ‘PL 200,000’ should not hide whether the figure is a direct tally or a projection.

Three different quantities often appear together: original hatchery stock, quantity dispatched, and quantity observed after receiving or acclimation. Each occupies a different place in the process. A transfer can lose animals through handling, retention on screens or survival changes, but the same difference can arise purely from two biased counting procedures. This article uses hypothetical teaching data to trace the figures from sampled water to a defensible statement; it does not describe a real facility or certify a stocking count.

Start by reading the measurement unit and stage

Before calculating, find the species, developmental stage, container size, counting method and observation time. Penaeid postlarvae and freshwater prawn postlarvae can behave differently, and their sampling performance may depend on size, temperature and activity. A stock estimate without species and stage is incomplete. ‘PL12’ conventionally indicates a postlarval age designation in certain shrimp systems, but a manuscript must define its convention; it is not by itself an assurance of quality or survival.

A typical volumetric estimate has the form total animals equals sample animals divided by sample volume times total water volume. In symbolic terms, N equals c times V, where c is animals per millilitre and V is the container’s millilitres. The formula does not establish that c represents the container. The critical physical requirement is that the aliquot’s animal density is representative at the moment of collection.

Every figure needs its own unit. A ten-millilitre tube is a volume, not a density. A count of fifty animals is a count, not fifty animals per millilitre. A one-litre holding vessel equals one thousand millilitres. Confusing those facts generates errors by factors of ten or one thousand. The translated report should place units inside column headings and show the exact conversion that led to the final stock total.

The label ‘sample’ may mean a withdrawal from a vessel, a net sweep, a whole transport bag, or a fixed number of animals chosen for examination. Only a volumetric withdrawal supports straightforward multiplication by a water-volume ratio. A net catch is subject to capture efficiency, while a biological quality sample may deliberately oversample weak or abnormal individuals. Translating both as ‘random sample’ can falsely imply equivalent designs.

Worked example: convert an aliquot to the estimated tank population

Suppose a mixed receiving vessel contains exactly 24 litres of water. Five separate 100-millilitre samples contain 94, 101, 97, 104 and 99 postlarvae. Their sum is 495; divide by five to obtain a mean of 99 animals per 100 mL. That is 0.99 PL/mL. Convert twenty-four litres to 24,000 mL, then multiply 0.99 by 24,000. The result is an estimated 23,760 postlarvae in the vessel.

The same calculation can be done without first converting density to animals per millilitre: 24,000 mL divided by 100 mL per sample gives 240 equivalent sample volumes. Multiply by the mean count of ninety-nine to get 23,760. Keeping the two paths in a training record is useful because agreement tests the arithmetic. It does not independently test sampling bias, since both paths use the same five observations.

Notice that ‘23,760’ is more digits than the biology may justify. The vessel could be slightly underfilled, the aliquots may not be independent and postlarvae can move while sampling occurs. Reporting approximately 23,800 PL, with the exact calculation retained in a method note, is often more defensible. If the business transaction requires an agreed numerical convention, document how estimates are rounded and how disputes or recounts are handled.

What if someone accidentally treats the sample as 10 mL instead of 100 mL? They calculate 9.9 PL/mL and estimate 237,600 animals: a tenfold error. A useful audit check is to reconstruct the sample jar and inspect the pipette or measuring cup used. Dimensional analysis can identify inconsistent units, but the original physical record must resolve an incorrectly transcribed sample size.

The volume problem: gross water is not necessarily sampleable water

Containers may have nominal capacities, fill levels, foam, screens, water exchanged during sampling or dead zones that affect representativeness. A labelled thirty-litre drum may hold twenty-six litres of actual liquid at the time of measurement. In a calculation at 1.2 PL/mL, the nominal volume predicts 36,000 animals, while twenty-six litres predicts 31,200. The difference is 4,800 animals even before considering sample variation.

If a hatchery concentrates animals into a smaller volume before counting, identify which volume was measured. Suppose larvae from forty litres are collected into four litres and three samples from that concentrated fluid give two thousand animals per litre. The total in the concentrate is eight thousand, not eighty thousand. Multiplying by the original forty-litre volume after concentration applies the concentration twice and overstates abundance by a factor of ten.

Conversely, an assay that dilutes crowded sample water must reverse the dilution. A ten-millilitre original sample mixed with ninety millilitres of clean water is diluted tenfold. If a one-millilitre subsample from the diluted mixture contains an average of eight larvae, estimated original density is eighty larvae per millilitre, assuming complete mixing and no losses. The calculator can perform the multiplication, but only the procedure log tells us whether the tenfold factor is correct.

Volumes change during transfer if rinse water is added, bags are opened or water is discarded. Each independent count must use the volume of its own counting container at its own time. Taking yesterday’s vessel volume and today’s concentration is not a legitimate estimate merely because both numbers appear on the same sheet. A translation should retain timestamps next to volume and sample figures whenever the process involves changing water levels.

Why swimming behaviour can bias volumetric samples

Postlarvae do not behave like dissolved salt. They can swim, react to light, avoid or approach flow, cling to surfaces or distribute unevenly when aeration changes. An aliquot taken from the surface after a prolonged settling period might differ markedly from one collected while the vessel is gently mixed. The arithmetic assumes each millilitre represents the average density, which is a biological and hydrodynamic assumption rather than a mathematical truth.

A published experimental study of postlarval penaeid shrimp counting found that volumetric subsampling could underestimate real populations, with performance changing under different size and temperature combinations. Its specific percentages describe that experimental system, not every modern shrimp species, tank or sampling device. The general lesson is that counting bias must be tested against known populations under relevant conditions rather than assumed away because a calculation is convenient.

Some operators respond to variation by mixing more vigorously, but excessively strong agitation can harm delicate animals or create an unsafe handling environment. A defensible approach is to use a validated, species- and stage-appropriate handling method that creates sufficient uniformity while protecting welfare. Recording the mixing duration, location of sampling and time between samples makes future disagreements diagnosable. The method must be part of the result.

If counts change systematically as sampling progresses, examine whether the container is losing volume, animals are aggregating, the first collection disturbed the group or the sampler became more effective with practice. Plot aliquot count against sample order instead of merely computing a final average. A trend does not establish its own cause, but it warns against treating observations as interchangeable replicates.

Replicates: the difference between counting more and sampling better

Five repeats from the same spot are not necessarily five independent views of a tank. They may be repeated measurements of the same local patch, particularly if the fluid is poorly mixed. To strengthen an estimate, define the sampling plan in advance: where aliquots are collected, how the vessel is mixed, whether sampling intervals are spaced and how each replicate is labelled. Repetition becomes valuable only when it probes the uncertainty relevant to the decision.

Consider counts per 100 mL of 80, 81, 79, 82 and 78. They appear highly consistent, but a surface-only protocol could still be biased low if larvae are concentrated near the bottom. A second plan that samples across depths might show 60, 82, 110, 93 and 105. Its higher variability is not necessarily a worse measurement; it may reveal spatial heterogeneity that the original plan hid. The question is whether the new design better reflects the actual container.

Random error can be summarised using a sample mean and dispersion. For five hypothetical counts 94, 101, 97, 104 and 99 per 100 mL, the mean is 99. The standard deviation, calculated with the usual sample denominator of four, is about 3.81 PL per 100 mL. That quantifies variation across these aliquots, but a confidence interval also requires assumptions about sampling design and independence. Never present a narrow interval as proof that aggregation bias is absent.

If the sampling procedure is not validated, use wording such as ‘volumetric estimate based on five aliquots’ and report the observed range. Do not transform standard deviation into an invented accuracy guarantee. Validation against a known count, such as controlled trials with a manageable number of postlarvae, is a separate exercise. Method verification is especially valuable when the estimate determines an expensive feed order or high stocking density.

Counting consistency and observer effects

Live, translucent larvae may overlap in a viewing dish, and technicians can interpret movement or fragments differently. Define what constitutes one count: intact mobile PL, all identifiable PL regardless of vitality, or a separately documented category of visibly dead animals. If the purpose is animals available for stocking, counting dead animals in the same numerator changes the meaning. Record these categories independently when feasible.

Observer agreement is another measurement property. Two technicians can count the same vial independently and compare their results before extrapolating to a large tank. If one regularly counts shells or detritus as PL, training and image examples may reduce disagreement. Averaging two numbers is not enough when the classification rules differ. The aim is shared criteria and verified technique, not an artificial consensus.

Device differences deserve attention. A large measuring cylinder, pipette and ladle do not necessarily withdraw the same type of sample. Small apertures may selectively exclude larger PL; a net may lose tiny animals; transfer surfaces may retain some. Even when each is marked ‘100 mL’, the sampled population can differ. The report should specify equipment and method rather than implying that equal nominal volumes mean equal sampling performance.

Time can also change the count. Moving animals from a bag into a holding tank may change their behaviour, and sampling immediately after a disturbance may produce different estimates from sampling after a standardised equilibration period. The timing should follow an appropriate husbandry protocol, not an invented universal waiting time. A translation that omits the sequence erases information needed to reproduce the result.

Uncertainty and the limits of precise-looking stock totals

The simple extrapolation gives one estimate, but the actual population may be larger or smaller. Major sources of uncertainty include volumetric calibration, unequal distribution, sample representativeness, individual counting errors, animal losses and time differences. A good reading separates uncertainty that can be estimated from replicates from bias that must be investigated experimentally. Treating both as a single percentage may conceal more than it clarifies.

Suppose the mean density estimate is 1,000 PL/L and the vessel volume is recorded as 20 L plus or minus 1 L. Even if density were known perfectly, volume uncertainty alone corresponds to roughly one thousand animals on either side of the central estimate of twenty thousand. If density is also uncertain, the total uncertainty is larger. A report that says ‘exactly 20,000’ is not supported by those inputs.

Do not combine percentage uncertainties by casual addition without specifying whether they are bounds, standard errors or independent random components. A worst-case bound can add relative maxima under conservative assumptions, while independent standard uncertainties may be combined differently. These are distinct questions. For an operational hatchery sheet it may be better to report the estimate, sample range and known systematic limitations than to manufacture a sophisticated statistical interval.

A useful practical sensitivity test is to recalculate the final count under plausible input changes: actual volume twenty rather than twenty-two litres, measured density nine hundred rather than one thousand PL/L, or a suspected low-biased sampling method. If these variations change a stocking decision, invest in a better measurement. Numerical sensitivity can tell an operator when the sampling procedure deserves greater attention.

Dispatch count and receiving count belong to different systems

A shipment might leave a hatchery labelled 120,000 PL and arrive with a receiver estimating 108,000. The difference is twelve thousand or ten percent of the dispatch claim, but it should not automatically be labelled ten percent mortality. The two estimates may use different mixing, volumes, observer rules or measurement times. True survival loss is only one of several possible explanations, and mortality requires its own evidence.

First reconcile the number of bags or compartments, their identifiers and the intended allocation. A lost bag is a logistics event, not a biological percentage calculation. Next compare methods, timing and counted organism categories. If both parties sampled the same population but with different biases, a discrepancy may exist even without any loss in animal numbers. A trustworthy report preserves the competing observations rather than choosing one silently.

Receiving quality indicators such as activity, temperature, salinity or visual condition may be relevant to husbandry decisions but should not be used as a mathematical substitute for the count. Similarly, a healthy-looking sample does not prove the whole shipment is disease-free. When health or transport concerns arise, follow established aquatic animal health protocols and obtain qualified assistance. A counting guide cannot determine quarantine status or certify infection absence.

Clear labels help commercial disputes. ‘Hatchery dispatch estimate, based on five 100 mL aliquots’ and ‘receiver estimate, based on three 50 mL aliquots’ tell the reader that the observations are not mechanically identical. Include whether the receiving vessel was remeasured after opening or dilution. Only after this reconciliation should the parties decide whether a recount, calibration trial or other verification is necessary.

Stocking density: converting numbers into animals per area or volume

When a farm uses the postlarval estimate to plan stocking density, a second denominator enters. A pond reported at 12,000 square metres is an area; a nursery tank at 15 cubic metres is a volume. The choice depends on the system and the intended stocking metric. If 240,000 PL enter a 12,000-square-metre pond, the initial arithmetic density is twenty PL per square metre. That is not twenty PL per cubic metre and not the number of surviving animals a month later.

Consider the volume variant. A tank with fifteen cubic metres equals fifteen thousand litres. Stocking 75,000 PL creates an initial count density of five PL per litre. The calculation alone does not certify that this is a safe or desirable husbandry density. Species, stage, oxygen supply, filtration, nutrition, biosecurity and equipment determine actual carrying capacity. Unit correctness is necessary, not sufficient, for a stocking decision.

If density is computed from an estimate, report the source of the numerator. A statement such as ‘initial stocking density approximately 20 PL/m² based on volumetric dispatch estimate’ is more honest than ‘actual density 20 PL/m²’. The former acknowledges uncertainty. If an official inventory later revises the initial count, the density should be recalculated with a dated note rather than silently overwriting the original stocking record.

Do not divide a pooled number by one compartment’s area. If six ponds each receive an estimated thirty thousand PL, total stock is 180,000 but an individual pond’s density requires its own measured area and allocation. The correct pooled density across all ponds uses total count divided by total relevant area, whereas management may require pond-by-pond densities. An average cannot reveal an overloaded pond when allocations are unequal.

A transfer-and-recovery scenario that exposes the denominator trap

Imagine a nursery dispatches an estimated 50,000 PL in two bags, nominally twenty-five thousand each. The receiver transfers both bags into a 10 L counting container after adjusting water according to an approved acclimation procedure. Three 20 mL aliquots yield counts of 94, 106 and 100. The mean is one hundred PL per 20 mL, or five PL/mL. Multiplying by ten thousand millilitres gives an estimate of 50,000 PL present at the receiving count.

Now suppose the counting vessel actually contains nine litres, not ten. Density is still estimated at five PL/mL, but total becomes 45,000. A reported ninety percent recovery appears without any change in the aliquot counts. The apparent ten percent discrepancy was caused entirely by a volume correction. It would be wrong to describe the missing five thousand as confirmed deaths. The revised count is a better estimate under the corrected volume, not a direct mortality measurement.

Alternatively, suppose the container truly contains ten litres but the sampling protocol misses larvae collected near a bottom screen. Correcting that bias requires evidence, such as a validated mixing test or a known-count recovery experiment. Simply multiplying the total by 1.1 because the operator expects a higher number replaces evidence with a preference. The purpose of the estimate is to learn about the stock, not to make the dispatch and receiving ledgers agree.

One practical communication method is to preserve a three-column account: hatchery dispatch estimate, independently sampled receiving estimate, and directly observed dead or retained animals where those are separately measured. These columns answer different questions. Their difference is a reconciliation problem, not automatically a biological survival rate. This is especially important when the same spreadsheet later feeds mortality, growth and feed conversion calculations.

Why a mass-based proxy needs its own calibration

Some facilities use weight or bulk volume after first counting a smaller reference batch. For example, an operator counts 500 PL into a container and weighs their combined drained wet mass under a specified method. If that reference mass is 2.0 g, a second identically prepared group weighing 20 g might be estimated at 5,000 PL. The extrapolation assumes comparable individual masses, draining procedure and equipment. It is not a direct count.

Postlarval size distributions can change with stage and cohort. A later batch of larger animals could weigh more per individual, making the same grams correspond to fewer PL. Water retained on nets or dishes can also dominate very small biological masses. The method should be calibrated on the current population and checked against independent samples. A weight estimate derived from last month’s cohort should not be copied into today’s report without validation.

A volume-based proxy likewise requires calibration. A ladle that visually resembles a counted reference ladle may hold a different amount of water or a differently concentrated group when transferred. The FAO freshwater prawn guidance describes practical approximate transfer counting methods while acknowledging their limitations. Such operational methods can be useful, but the translation must identify them as estimates and not invent laboratory-level exactness.

If a buyer’s contract depends on a fixed number, the sampling method and acceptable tolerance should be agreed in advance. This is a measurement-governance issue, not merely a mathematical one. A strong record makes transparent whether numbers were obtained by individual count, volumetric subsampling, weight proxy or visually calibrated batches. Each method has different strengths, failure modes and appropriate claims.

A complete laboratory-style example with dilution and sampling

Consider a transfer sample from a larger nursery vessel. A technician takes 25 mL of well-mixed original water and adds 75 mL of clean water, yielding a final dilution volume of 100 mL. The dilution factor is four, because one original part becomes four total parts. A 10 mL aliquot of that diluted mixture contains 22 PL. Its measured diluted density is 2.2 PL/mL, so the inferred original density is 8.8 PL/mL.

If the original vessel contains 5.0 L, or 5,000 mL, the estimated population is 8.8 times 5,000, or 44,000 PL. Check by another route: the 25 mL original sample is estimated to contain 220 PL, since the 100 mL dilution contains 2.2 PL/mL. There are 5,000 divided by 25, or 200 such original sample volumes. Multiplying 220 by 200 again produces 44,000.

Now inspect uncertainty and practical design. The count of twenty-two in the diluted ten-millilitre aliquot is relatively small. A second aliquot might contain nineteen or twenty-six purely from sampling variation, and poor mixing might create larger differences. The observed count should not be promoted into an exact whole-tank census. Multiple independent aliquots and a clear method can improve the estimate, but each additional sample should be recorded before results are combined.

Imagine the original 25 mL sample was accidentally topped up to 125 mL rather than 100 mL. The true factor is five, not four. With the same diluted density of 2.2 PL/mL, the corrected original density is 11 PL/mL and estimated vessel total becomes 55,000. This example illustrates why the final dilution volume, rather than a casual description such as ‘diluted with water’, must appear on the record.

Comparing lots: the denominator changes again

Suppose Hatchery A sells a batch estimated at 1.2 million PL, of which 1.0 million enter a receiving system by volumetric estimate. Hatchery B sends 0.8 million, of which 0.72 million are estimated at reception. The reported receiving-to-dispatch ratios are about 83.3% and 90%. It is tempting to call Hatchery B more successful, but that conclusion requires comparable measurement protocols and evidence about actual loss, not just two inferred totals.

If Hatchery A used a sampling device with a known low-capture bias, its receiving count may be depressed even when biological recovery is good. If Hatchery B counted earlier and before handling losses, its ratio may be overstated relative to the final stocked number. Commercial performance metrics must align the endpoints. The same rule applies to the labels stocked, delivered, survived and harvested, each of which belongs to a different stage of production.

Mean size also matters when comparing counts and biomass. One million small PL may weigh substantially less than one million older PL. Count comparisons should not automatically be used to infer equivalent nutritional demand or handling risk. When a report includes both counts and masses, state separate units and the procedure used to obtain each; do not turn one into the other using an unverified universal average weight.

A careful conclusion names what the evidence supports. It may say that two suppliers’ records show different reported postlarval counts, and that differences in protocol prevent a reliable performance ranking. This is not evasive language. It is the correct recognition that an apparent statistical difference can be produced by measurement design. A valid comparison begins with harmonised stages, sample times and denominators.

Data cleaning and translation across languages

In a bilingual production spreadsheet, PL can be confused with a unit of volume or an undefined life stage if the abbreviation is never expanded. Translate it as postlarvae only after confirming the biological context, and retain any stage numbering convention alongside it. Separate total animals, animals per litre, animals per bag, and live animals recovered; short headings like ‘number’ do not tell the reader which quantity is meant.

Numerical formats create additional hazards. A column with entries ‘1.250’ and ‘1,250’ may use regional punctuation conventions differently. Determine whether the values represent one point two five or one thousand two hundred fifty from the original locale, units and linked calculations. A translated number should retain its quantitative meaning, not merely its visible characters. Where verification is impossible, the ambiguity should be noted rather than silently normalised.

Use plain-language descriptions of the method: ‘The operator stirred the container, withdrew five 100 mL aliquots, counted postlarvae and extrapolated to the measured 24 L volume.’ This sentence tells more than an unexplained equation and is easier for a buyer to audit. Technical accuracy does not require dense prose; it requires preserving the relationship between observation, denominator and inference.

Terms such as estimated, counted, observed, assumed, calibrated and verified must remain distinct through translation. An estimated stock size derived from aliquots is not an observed census. A calibrated ladle is not automatically validated across all postlarval stages. A verified count method has evidence for the relevant conditions, not simply a written procedure. These words indicate different levels and kinds of evidence and should be used consistently.

Errors and repairs: a field audit sequence

Begin with arithmetic and units. Check that litres were converted to millilitres, that sample volumes match the counting vessel and that all dilution or concentration steps are accounted for exactly once. Recalculate one full example by a second algebraic route. This catches copy errors without requiring fresh biological measurements. If a spreadsheet formula refers to the wrong cell, fix the formula and preserve the original and corrected result with a dated note.

Next inspect physical evidence. Compare fill markings with actual water volume, look for settling or aggregation, and confirm how the sampler operated. Review differences between replicate positions and between observers. If estimates differ by a factor of ten, search for dilution or unit mistakes first; if they drift gradually across samples, examine distribution, sampling order and changing volume. These are diagnostic clues, not proof of a particular cause.

Then reconcile the business and biological ledgers. Check bags dispatched, bags received, volumes after mixing, observed dead animals where reliably counted and the point at which the stock entered the nursery. Do not infer mortality merely because dispatch and receipt estimates differ. When a material discrepancy remains, record it and use an agreed independent verification process rather than revising one estimate until it matches another.

Finally, review whether the count was adequate for its decision. A rough estimate might support provisional preparation, while a high-cost transfer may justify more elaborate validation. The appropriate level of effort depends on consequences, animal welfare and the operating system. There is no universal sample count that produces a universally accurate population figure. Evidence quality comes from the whole design, not one number of replicate aliquots.

Four transfer exercises and their answers

Exercise one: A 12 L vessel produces four 50 mL counts of 58, 62, 61 and 59 PL. The mean is sixty per 50 mL, or 1.2 PL/mL. Twelve litres is twelve thousand millilitres, so total estimated population is 14,400 PL. If the vessel holds only eleven litres, the revised estimate is 13,200. State both the observation and the volume assumption; an estimate changes when its measured basis changes.

Exercise two: A concentrated 2 L holding vessel contains animals taken from a 20 L tank. Three undiluted one-millilitre counts average 250 PL. Estimated count in the concentrate is 250 times 2,000, or 500,000 PL. Multiplying by the original 20 L would give five million, an error caused by ignoring that the animals have already been concentrated. The correct denominator is the actual counted vessel volume.

Exercise three: A 20 mL original sample is diluted to 100 mL. A five-millilitre diluted aliquot averages thirty PL. The diluted density is six PL/mL; the original density is thirty PL/mL because the dilution factor is five. For an original vessel of 4 L, the estimated population is 120,000 PL. Check by reconstructing the diluted sample: one hundred millilitres at six PL/mL holds six hundred animals from the twenty-millilitre original aliquot.

Exercise four: Two dispatch records each claim 100,000 PL. One receiving count estimates 92,000, the other 87,000. The mathematical differences are eight and thirteen percent of dispatch estimates. Are those verified mortality rates? No. The data do not show deaths; the estimates may differ in sampling design, volume or timing. A correct answer names the observed discrepancies and states what additional evidence would be needed to interpret biological loss.

Frequently asked questions

Can I treat a count from a small cup as the exact number in a pond? No. It is an extrapolation and needs a representative sampling design. Animals do not remain uniformly distributed simply because the water looks stirred. In a large pond the problem becomes substantially more complicated, so a small-cup count cannot automatically represent its entire post-stocking population.

What if replicate counts agree almost perfectly? Agreement is useful evidence of repeatability, but it cannot exclude a consistent low-capture or sampling-position bias. A method that misses half the animals on every attempt can be extremely precise and badly inaccurate. Independent validation against known counts under comparable conditions is the appropriate test when accuracy matters.

Do I need to count every animal to plan stocking? Not always. Volumetric or other calibrated estimation methods are common because direct census can be impractical. The appropriate choice depends on animal welfare, operating constraints, financial consequences and the required accuracy. What matters is identifying the method as an estimate and documenting its limitations instead of treating it as a perfect count.

Why is the postlarval stage important to the number? Size and behaviour can affect sampling, and a method calibrated on one stage may not perform the same way on another. Stage also changes husbandry needs. The label therefore belongs in the method record. A stage designation alone, however, cannot prove health or establish that every animal will survive transfer.

How should a translated invoice display the count? A useful description is ‘Estimated 120,000 postlarvae, based on volumetric subsampling at dispatch; method and receiving verification recorded separately.’ If the contract establishes another wording or formal counting standard, follow it faithfully. Do not invent regulatory guarantees, precision or an independently verified census that the source documentation never supplied.

What is the first check when stock numbers are ten times apart? Verify millilitre/litre conversion, withdrawal volume, dilution and concentration factors, and current container volume. Large powers-of-ten discrepancies often arise in unit bookkeeping. If those checks pass, inspect sample representativeness and the actual counting procedure. The arithmetic should lead the investigation, not end it.