You can do everything right at the bench, clean washes, careful timing, consistent pipetting, and still feel stuck when the plate reader delivers a grid of uncrunched numbers.
That’s because the ELISA output does not provide concentrations per se, the plate reader provides optical density values that need to be interpreted. The decisions you make between output and analysis (background correction, curve fitting, dilution factors, and basic quality checks) are where false positives, inflated values, and hard to reproduce results most often creep in.
Below is a practical, start to finish workflow to guide you through. It applies to most ELISA kits and in house assays, with a few notes for common format differences.
Step 1: Confirm the kind of ELISA you are running
Before you touch the spreadsheet, be clear on the goal.
Quantitative analysis requires a standard curve and interpolation. Qualitative and semi quantitative assays still need controls and a defined decision rule for calling a sample positive or negative.
Also confirm the assay format (sandwich vs competitive). Competitive ELISAs often produce an inverse curve where signal decreases as concentration increases, which affects curve fitting and interpretation.
Step 2: Export and document the raw plate readings
For TMB based assays, the primary read is commonly 450 nm after adding stop solution. Many protocols also use a reference wavelength (often 570 nm, 620 nm, or similar) to correct for optical imperfections in the plate. If your protocol specifies a reference read, apply it the same way every time.
Timing matters. Read within the time window specified in your protocol, and keep the delay between stopping and reading consistent across plates.
Export the raw OD values exactly as read by the instrument and save them as an untouched file.
Step 3: Apply background correction thoughtfully
Most assays include one or more “zero” wells, often called blanks or zero standard wells. These are treated exactly the same, aside from analyte addition.
A common approach to background correction is:
This corrects for reagent and instrument background. It does not fix nonspecific binding by itself, which is why negative controls and matrix controls also matter.
Tip: Be consistent about which wells you use for background. Some labs use the zero standard rather than an empty well because it better matches the sample diluent matrix.
Step 4: Build the standard curve using the right model
Across the full concentration range, many ELISA standard curves are not linear. They typically show a lower baseline, a rising region, and an upper plateau.
Because of that, 4 parameter logistic (4PL) fitting is commonly used. 5PL can be helpful when the curve is noticeably asymmetric.
Choose a model based on the curve shape and your lab’s standard practice, then stick with it within a study. Tools like GraphPad Prism, SoftMax Pro, and other ELISA analysis software can fit 4PL and 5PL curves and report curve parameters.
A practical quality check is not just “does it look smooth,” but whether:
Step 5: Interpolate unknowns only within the quantifiable range
Use the fitted curve to interpolate sample concentrations.
Only trust results when the sample OD falls within the portion of the curve where the assay can quantify accurately. If a sample is above the upper end, dilute and rerun. If it is near background, you may be below the lower limit of quantification.
A good habit is to flag samples as:
This prevents accidental over interpretation of numbers that the assay cannot support.
Step 6: Correct for dilution factors
If you diluted your sample to bring it into range, apply the dilution factor after interpolation.
Example: If a sample was diluted 1 to 10, multiply the interpolated concentration by 10 to report the concentration in the original sample.
Also document any additional factors that change the final value, such as sample volume changes during processing, or unit conversions.
Step 7: Evaluate replicates using mean, SD, and CV
For each sample and standard, calculate:
High CV is often a sign of a technical issue such as inconsistent pipetting, incomplete mixing, uneven washing, or plate handling differences.
Many research workflows aim for CV values around 20% or lower, with tighter targets for higher confidence applications, but the right threshold depends on your assay, your lab standards, and the decision you are making from the data.
If duplicates disagree, do not automatically average and move on. First check for obvious causes like bubbles, edge wells, or a visible pipetting error.
Step 8: Confirm controls and scan for plate effects
Controls are not optional. They tell you whether the plate behaved the way you think it did.
Check that:
Edge effects can happen due to evaporation and temperature differences during incubation. If you see systematically higher or lower signals on the outer wells, consider using plate seals, controlling incubation conditions, or avoiding edge wells for critical samples.
Step 9: Normalize only when it helps, then compare groups carefully
Normalization can be useful when you are comparing across plates, days, operators, or reagent lots. Common strategies include using a shared inter plate control sample, or normalizing to a baseline control group within each plate.
If your standards and controls are stable and you are reporting absolute concentrations, you may not need additional normalization. The key is to avoid mixing approaches within the same analysis.
Once your data are in a reliable form, choose statistical tests that match your design and assumptions (for example, whether groups are paired, whether variance is similar, and whether transformation is appropriate). Statistics cannot rescue unstable assay performance, so do those critical quality checks first.
A final note
Reliable ELISA results are built as much at the keyboard as they are at the bench. If you save the raw OD file, apply background correction consistently, fit an appropriate curve, and flag anything outside the quantifiable range, your numbers become defensible, repeatable, and easier to troubleshoot later. The goal is not a perfect looking curve. The goal is knowing which results you can trust, which ones need a dilution and rerun, and why.
If you want a simple way to stay consistent across plates and projects, use a checklist and treat it like part of the protocol, not an optional afterthought.