ELISA workflow reference
Within-Run and Between-Run Precision in ELISA Workflows
Precision is not a single property printed at the end of an experiment. It is a record of how much repeated measurements vary under defined conditions—and whether those conditions remain comparable from plate to plate, day to day, operator to operator, and lot to lot.
In this guide
- Define the precision question
- Within-run precision
- Between-run precision
- Controls and plate planning
- Calibration and curve review
- Matrix effects and interference
- Species-specific selection
- Documentation and interpretation boundaries
01 / The question behind the number
Precision begins with a defined source of variation
In an ELISA workflow, repeated results can differ even when the underlying sample or experimental condition has not changed. Pipetting, incubation timing, washing, plate position, reagent preparation, calibration, reader settings, analyst technique, storage history, and sample matrix can all contribute to observed variation. A useful precision assessment therefore starts by stating what is being held constant and what is deliberately allowed to change.
Within-run precision describes variation among replicate measurements made within one defined analytical run. The phrase “same run” should have operational meaning: the same plate or coordinated plate set, reagent preparations, instrument session, operator or procedure, and timing plan, as applicable to the study. Between-run precision describes variation when the measurement is repeated across separate runs. Those runs may differ by day, analyst, reagent preparation, plate, instrument session, or other factors selected in the study design.
These terms are related but answer different questions. Within-run variation asks whether the workflow is internally consistent under a controlled set of conditions. Between-run variation asks whether the workflow remains comparable when the run is repeated. A small first number does not automatically establish a small second number: repeating several wells on one plate does not test day-to-day drift, new standard preparation, a different operator, or a new reagent lot.
Start with a variation map
Before calculating a summary statistic, list the sources of variation relevant to the decision: replicate wells, plate, day, operator, reagent preparation, lot, instrument, dilution scheme, and sample matrix. Then label each factor as fixed, randomized, blocked, or intentionally varied.
This framing is especially important when comparing records in a catalogue. A product name or target label alone does not establish species suitability, matrix suitability, curve performance, or precision in a particular laboratory. Begin with the source documentation and assay context. The ELISA assay selection guide provides a starting framework for checking analyte identity, species, documentation, and intended research context before a kit enters a precision study.
02 / Same-run consistency
Within-run precision: what replicate wells can and cannot show
A within-run assessment commonly uses repeated measurements of one or more control materials, standards, or samples on the same run. The goal is not merely to obtain several numbers; it is to establish whether those numbers were generated under a clearly recorded common condition. Replicates are most informative when their placement, preparation, and handling are documented rather than treated as interchangeable wells.
Replicates are a design choice
Technical replicates can reveal local pipetting variation, edge effects, inadequate mixing, inconsistent washing, or other same-run problems. Their value depends on how they are distributed. Adjacent replicates may expose repeatability of a local operation but may not reveal a position-related pattern across the plate. A deliberately distributed plate map can help distinguish a well-specific problem from a broader plate effect, provided the distribution does not conflict with the assay procedure.
Replicate count should be justified by the purpose of the experiment and the available material. More wells do not repair an uncontrolled workflow. If all replicates were prepared from one poorly mixed dilution, they can be tightly grouped around a biased value. Conversely, a wider spread may reflect a true preparation or matrix issue rather than random pipetting alone.
Summaries should retain the underlying context
For measurements on a suitable scale, analysts may summarize replicate results with a mean and standard deviation, then express relative dispersion using a coefficient of variation when that is appropriate for the scale and data behavior. The calculation is not the interpretation. A relative measure can become unstable when the mean is close to zero, and transformed or bounded data may require a different summary. Report the number of replicates, the calculation convention, any exclusions, and the scale used.
Within-run precision should be read alongside the standard curve, control behavior, blank response, and plate map. A narrow cluster of replicate values on a visibly unsuitable curve is not persuasive evidence of quantitative reliability. Conversely, an isolated outlier should not be removed solely because it increases the spread. Review the well image or readout context if available, pipetting record, wash sequence, sample preparation, and any documented procedural deviation before deciding how to handle it.
03 / Repetition across conditions
Between-run precision: reproducibility across days, plates, and preparations
Between-run precision is a study of comparability. Each run should be treated as a complete measurement event with its own date, plate identity, reagent preparation, calibration record, analyst or operator record, instrument context, and control results. The exact factors to vary depend on the question. If the objective is routine repeatability, the study may reproduce the intended laboratory workflow. If the objective is robustness, selected factors can be varied deliberately.
A practical design separates run identity from sample identity. The same control or retained sample can be used across runs to provide an anchor, while study samples can be randomized or blocked to avoid confounding sample order with day or operator. If every sample is measured on only one day, a day effect cannot be distinguished from a sample effect. If every plate uses a different dilution scheme, dilution and plate effects become difficult to separate.
| Field group | Examples to record | Why it matters |
|---|---|---|
| Run identity | Date, plate ID, analyst, instrument session, protocol version | Defines the comparison unit and makes repeated runs traceable. |
| Material identity | Sample ID, control ID, dilution, preparation time, freeze–thaw history when relevant | Separates biological or material differences from process variation. |
| Reagent state | Lot or batch identifiers, preparation record, storage and expiry checks | Allows lot-associated shifts to be investigated rather than guessed. |
| Plate execution | Map, replicate positions, incubation times, wash record, deviations | Connects a result to the physical workflow that produced it. |
| Readout and review | Raw signal, blank, standards, calculated result, curve model, residual review, exclusions | Preserves the path from observation to reported value. |
When multiple runs are compared, inspect both the run-level summaries and the individual observations. A pooled average can conceal a shift affecting one entire plate. Plotting control results by run, for example, can reveal a directional pattern that is invisible in one grand mean. A run effect may be related to reagent preparation, temperature, timing, plate position, instrument response, or sample handling; the plot identifies a pattern, not its cause.
Lot considerations belong in the design rather than in a post hoc explanation. If a new lot is introduced between two groups, lot and group are confounded. A bridging plan—using suitable shared controls or reference material across the transition—can make a comparison more interpretable, but the plan and its acceptance logic should be defined before the results are reviewed. Do not infer lot equivalence from catalogue naming or from a single matching result.
04 / Controls make variation visible
Controls and plate planning: protect interpretability before the read
Controls do not all answer the same question. A blank can help characterize background from the assay system or plate process. Standards support calibration. A positive control, where appropriate and defined by the assay documentation, can indicate that the detection workflow produced an expected type of response. A negative control can help examine nonspecific or background behavior. Matrix controls or dilution controls may be needed when the sample environment itself is part of the uncertainty.
The appropriate control set depends on the assay design and research question. A control is useful only when its identity, preparation, placement, expected role, and review rule are understood. The control plan should also specify what happens when a control is missing, contaminated, outside a pre-defined range, or inconsistent with the rest of the plate. “Pass” and “fail” labels without an operational definition make later comparisons difficult.
Plate maps are records, not decoration
A plate map should preserve the relationship between well position and material. It should show standards, blanks, controls, samples, dilutions, and technical replicates. If samples are randomized, the randomization approach should be retained. If edge wells are treated differently, that choice should be recorded. When a plate is divided among several operators or preparation times, those boundaries should be visible in the map or run record.
For repeated runs, keep the map structure sufficiently comparable to support review while avoiding a design that permanently associates one sample group with one position. A balanced layout can reduce confounding, but it does not eliminate the need to inspect position effects. Controls distributed across the plate may help identify gradients or localized problems; their placement should follow the validated or documented procedure for the assay.
Researchers building a broader quality framework may also find the companion article ELISA Controls Explained: Building an Interpretable Plate Plan useful for thinking through control roles and plate-level interpretation. It is a related learning resource, not a substitute for the source documentation of a specific assay.
05 / From signal to reported value
Calibration and curve review: precision is downstream of model choice
ELISA results are generally inferred from a relationship between observed signal and standards of known or assigned concentration. The calibration model determines how that relationship is represented and how unknown signals are converted into reported values. Model selection should therefore be tied to the observed standard response, the assay documentation, the concentration range used, and the purpose of the measurement.
Do not select a curve model because it is familiar or because it produces a desirable result. Review the standard responses, replicate spread, fitted-versus-observed behavior, residual pattern, weighting decision where relevant, and the region in which unknowns are interpolated. A model can fit the standards globally while representing one portion of the range poorly. Unknowns outside the supported range should not be made interpretable merely by extending the curve mathematically.
Inspect the standards
Confirm identity, dilution sequence, replicate structure, signal ordering, and any apparent saturation or irregular response.
Fit and compare appropriately
Use a model and weighting approach justified by the assay context; preserve the selected method and alternatives considered.
Review residuals and range
Look for systematic departures, high-leverage points, uneven spread, and unknowns near or beyond the usable boundaries.
Carry uncertainty into reporting
Flag dilution, extrapolation, failed controls, or curve limitations rather than presenting a calculated value without context.
Curve review and precision review should be connected but not conflated. A stable curve does not prove that sample preparation is reproducible. Conversely, variable standards may make a run unsuitable for interpreting sample replicates even when those sample replicates are close together. Preserve raw signal and calculated outputs so that a later review can distinguish instrument-level variation, curve-fitting variation, and sample-preparation variation.
For a complementary discussion of model review and reporting boundaries, see Evaluating ELISA Standard Curves Without Overstating the Data. The article emphasizes that a curve is evidence about a defined calibration context, not a universal guarantee of performance across samples or laboratories.
06 / When the sample changes the measurement
Matrix effects and interference: observations that call for further validation
A matrix effect occurs when components of the sample environment alter the measured response relative to the behavior expected from the calibration system. The source may be biological material, extraction components, buffers, salts, detergents, endogenous binding partners, or another constituent. The important point is conceptual: a standard prepared in one solution does not automatically reproduce the behavior of an analyte in every sample matrix.
Several observations can prompt further investigation. A sample may produce a nonparallel dilution series, unexpected recovery after a known addition, a result that changes disproportionately with dilution, or a signal that behaves differently from the standards. A high background, unusual blank response, inconsistent replicate pattern, or apparent inhibition can also justify review. None of these observations alone identifies the mechanism. They indicate that the current evidence may not support a simple interpretation without additional work.
Dilution, spike recovery, and parallelism are questions—not badges
- Dilution: Does bringing a sample into a different concentration range change the result in a way that is consistent with the assay’s documented behavior?
- Spike recovery: When a known amount is added to a sample, is the observed change compatible with the intended measurement context and the pre-defined review approach?
- Parallelism: Does the dilution-response relationship resemble the relationship represented by the standards after accounting for the assay’s scale and model?
- Interference checks: Could a component alter binding, detection, background, or signal generation independently of the analyte concentration?
These experiments require explicit preparation details and acceptance logic. Recovery outside an expected range is not automatically proof of analyte absence or assay failure; it may indicate dilution error, instability, matrix interaction, incorrect assumptions about the added material, or another issue. Likewise, apparent parallelism does not establish that every sample type is validated. Report what was tested, under which conditions, and what remains unknown.
Matrix behavior can also affect precision estimates. If the same sample is diluted differently across runs, the observed between-run spread includes the dilution decision. If only one matrix is represented in the control material, its precision may not transfer to a different matrix. Track sample type, preparation, dilution, and any treatment that could change the measurement environment.
07 / Selection before comparison
Species-specific assay selection: similar names are not equivalent evidence
Species selection is a separate verification step from precision assessment. An analyte may have related names across species, but a familiar target label does not establish that the same reagents, calibrators, epitope, sample matrix, or biological interpretation apply across species. Confirm the species named in the product record and source documentation, the analyte identity or synonym context, the documented sample types, and any stated cross-reactivity or specificity information before designing a comparison.
This matters when a project compares measurements across humans, companion animals, rodents, livestock, or other organisms. A between-run study can be technically consistent and still answer the wrong biological question if the selected assay does not provide evidence for the intended species. Conversely, a product record that names a species and target is catalogue evidence for navigation, not independent evidence that the product is interchangeable with another record.
Verify the target
Record the analyte name, synonyms, species context, and whether the research question concerns concentration, presence, or relative change.
Verify the documentation
Check the source record and supplied technical information for sample type, assay format, calibration context, and stated specificity information.
Verify the comparison
Decide whether results are being compared within one species, across species, across products, or across runs—and do not treat those as the same claim.
For catalogue navigation, the supplied records illustrate two distinct paths: Dog NGAL ELISA Kit (catalogue number E-40NGL; target listed as NGAL (Lipocalin-2); application listed as ELISA) and Human CRP ELISA Kit (catalogue number E-80CRP; target listed as CRP; application listed as ELISA). These fields identify the records and their listed catalogue descriptors. They do not establish comparative performance, cross-species suitability, interchangeability, or endorsement.
Researchers can also browse the animal ELISA kits and human ELISA kits category pathways, then return to the searchable Products page to verify current records and narrow the catalogue search. Selection should remain anchored to the source documentation and the actual study design.
08 / Reporting what the experiment can support
Documentation and interpretation boundaries
A precision claim is only as transparent as its run history. At minimum, preserve the raw readout, plate map, standard and control identities, sample and dilution records, run date, operator or analyst record, reagent and lot identifiers when relevant, instrument context, curve model, calculation method, exclusions, deviations, and review decisions. Store enough information to reconstruct which observations were included in each summary.
A compact reporting checklist
- Define the unit: State whether the analysis concerns wells, prepared dilutions, samples, plates, or complete runs.
- Separate levels of variation: Do not label technical replicate spread as between-run reproducibility.
- Show the design: Identify which factors were held constant and which were varied across runs.
- Report the raw context: Include replicate count, controls, standards, curve review, and relevant exclusions.
- Investigate patterns: Review run-level shifts, plate position, lot changes, matrix behavior, and operator or timing differences.
- State boundaries: Identify unsupported extrapolation, untested species or matrix, missing controls, and unresolved deviations.
- Distinguish evidence from inference: A catalogue descriptor, a technical record, and an in-house precision result support different kinds of conclusions.
The most useful conclusion is often narrower than “the assay is precise.” A defensible statement might specify the material, concentration range, replicate structure, runs, factors varied, summary method, and observed limitations. For example, a report may say that replicate measurements of a defined control were summarized within a stated run structure and that selected runs showed a particular pattern of variation, while noting that the experiment did not evaluate another species, matrix, lot, or concentration range. That level of specificity makes the result easier to reproduce and harder to overread.
Precision is therefore a workflow property observed under defined conditions, not a permanent label attached to a target name. Use the searchable catalogue to locate candidate records, use source documentation to check fit, and use a planned, traceable experiment to establish what repeated measurements mean in your laboratory.
Related reading: A Practical Guide to ELISA Data Quality and Transparent Reporting extends the documentation perspective to broader reporting decisions.
This article explains experimental reasoning and catalogue navigation. It does not provide product-specific precision claims, establish clinical performance, or replace the technical documentation and validation plan applicable to a particular assay.