Stop Blaming Yourself: 3 Troubleshooting Tools For When Experiments Go Wrong

Ever re-run the same experiment three times just to keep getting “bad” results? In reality, no result is “bad” and experiments never “fail”. Instead, they produce data; sometimes that data is clear, sometimes it is not.

After running an experiment, your results are either conclusive or inconclusive, and an inconclusive result is not useless. If you treat inconclusive results as a “failure”, you miss the chance to discover what actually went wrong and learn from it.

What does “inconclusive” actually mean in practice?

An experiment is inconclusive when it does not clearly support or reject your hypothesis. This can look different depending on the assay, but common indicators include:

• Replicate variability that exceeds your assay’s accepted coefficient of variation
• Controls that do not behave as expected (e.g., positive controls with no signal, or negative controls with background above threshold)
• A signal-to-noise ratio too low to distinguish any real biological effect from artifacts
• Unexpected dose–response curve shapes, such as non-monotonic behavior in a titration
• Results that conflict with well-established internal benchmarks without an obvious technical explanation

Recognizing your inconclusive result as more than “bad data” is the first step towards getting clarity. The next step is to troubleshoot and determine whether the data supports or rejects your hypothesis.

Put this article into practice

Choose a free resource to help you move forward

EBOOK

Guide to Lab Safety

This ebook introduces you to some of the most common safety principles that you will encounter every time you step into a laboratory. On top of that, it will help point you towards additional resources to answer your questions when deeper questions arise.

GET YOUR COPY

CHEAT SHEET

Guide to Pipette Cleaning

Keeping your pipette clean and sparkling is critical. After all a contaminated pipette could ruin your precious samples in the lab! Download our cheat sheet to get a simple guide to cleaning your pipette. Stick it up in the lab to help remind yourself how to look after this vital lab tool.

GET YOUR COPY

Why Experiments Go Wrong In The First Place

How can you trust the observed results from an experiment? This comes down to two essential factors in scientific research: repeatability and reproducibility.

1. Repeatability: the closeness of agreement between results obtained under identical conditions, e.g., replicates within a single experiment.
2. Reproducibility: the consistency of results across changed conditions, such as different times, operators, calibrators, or laboratories.

The absence of either factor can undermine your data.

Repeatability VS Reproducibility

Checking the replicates first is a good starting point for most experiments. If they’re inconsistent, you’ve probably got a repeatability problem. If your replicates are consistent but cross-runs are not, you have narrowed the issue down to something that changes between sessions, e.g., a different reagent lot, a different operator, or some environmental variation.

However, it is important to note that “reproducible” does not mean correct! An experiment can return consistent results across multiple runs and still be systematically wrong. This is often seen in calibration drift. If a standard curve shifts and is not revalidated, that source of truth (your control) and every downstream measurement will be incorrect.

Biases like this are particularly dangerous because they do not trigger the random-error alarms of variability.  If your results can be reproduced but conflict with orthogonal evidence or positive controls, then consistency is a reason to dig deeper.

Recognizing Failure Patterns Before Root Cause Analysis

The entire experimental and testing process is far more complex than it appears. Vulnerabilities exist at every step along the chain, from sample collection to storage to preparation and handling, and finally measurement.

A problem at any one of these points can compromise your results, which is exactly why operator error is rarely a satisfying or useful explanation. Some common reasons experiments return inconclusive results are:

Rethink Blame: Practical Tips for Resolving Issues

When experimental results do not make sense, our instinct is to assume operator error. However, unclear blame is neither a diagnosis nor a fix. The ability to troubleshoot effectively depends on specific, repeatable behaviors that you can include in your process and written procedures:

1. Record any discrepancies as they take place. Any protocol deviations should be noted as they occur, rather than relying on memory. A deviation log is a useful troubleshooting tool that you will never regret keeping.
2. Never re-run an inconclusive experiment. If you repeat this same protocol without changing a single variable, you are likely to get the same result. A specific factor to focus on, control, vary, or monitor differently must be identified before repeating your experiment.
3. Record what you expected. Before analyzing results, write down what you expected to see. Compare this explicitly to what you actually observed. The gap between prediction and outcome is where diagnostic information lives.
4. Do not dismiss anomalies. An unexpected band, an outlier well, or a shifted baseline is usually due to a systemic effect. Recurrent experimentation anomalies usually indicate a specific issue with your protocol or materials.
5. Separate facts from your assumptions. While framing a problem, point out the facts (e.g., the positive control signal was 40% of expected) as opposed to the assumptions (e.g., the control antibody must have degraded). Base your actions on facts, and question any assumptions.

Laboratory errors are better understood as indicators of underlying weaknesses in procedures and systems, rather than as assumptions about what went wrong. Identifying those gaps is what allows you to resolve issues efficiently and prevent them from recurring, rather than simply repeating the experiment and hoping for a different outcome.

When to Troubleshoot VS When to Stop

Not every inconclusive result warrants a re-run. Before embarking on a root cause analysis, identify what solution is required:

1. Re-run: There may be times when you are confident in your protocol, and you can attribute the deviation to a single item (e.g., a broken incubator or an out-of-date reagent). In this case, it’s fine to re-run the assay and confirm the diagnosis. Change only the suspected variable and repeat the experiment.
2. Troubleshooting: When an unexpected result reoccurs across two or more independent runs, or when no single deviation can be identified, troubleshoot. Use the frameworks described in the following section, “Designing Experiments for Diagnosis,” to identify system-level causes before rerunning any experiment.
3. Re-design: If troubleshooting reveals that the assay cannot answer your question under current conditions, consider re-designing it. At this point, further repetition or root cause analysis is pointless, so reconsider your question and rework the experimental structure.

Designing Experiments for Diagnosis

Keep in mind that the best time to make an experiment diagnosable is before you run it. Make sure to add controls that independently test each critical assumption, not just the positive and negative endpoints. Be sure to note everything that could change during your run, such as reagent lots, instrument serial numbers, temperature, and who was performing the analysis.

And if possible, change only one variable at a time, or else you won’t know what is causing the problem. Experiments that are well-designed and documented can be diagnosed quickly and are far less likely to trap you in cycles of uninterpretable repetition.

Tool 1: The 5 Whys

In the 5 Whys framework, you start with your problem (e.g., the ELISA results weren’t consistent) and you keep asking “why?” until you get to a fundamental fixable answer. For example:

• What caused the ELISA results to differ among replicates? Low concentrations showed a non-linear standard curve.
• What caused the non-linear standard geometric curve? The assay’s detection range did not include the low concentrations.
• Why were the standards below the detection range? The standard stock had been degraded, lowering all concentrations.
• What caused the stock standard to degrade? It had several freeze-thaw cycles over several weeks without single-use aliquoting.
• Why was it not aliquoted? The lab’s SOP specified aliquoting, but the current batch was prepared by a new team member who had not been trained on that step.

The root cause here is a training gap, not a pipetting error or a “bad result.” Without this laddering, you might remake the standards and re-run the assay. With it, you fix the process that allowed degraded reagents to enter the workflow in the first place (Figure 1).

Limitation: The 5 Whys assumes a single linear causal chain. It can be misleading when the failure has multiple independent contributing factors, or when the initial problem statement is poorly defined. If you find yourself branching at a “why,” switch to a Fishbone diagram instead (next section).

Figure 1: Worked example of The 5 Whys adapted from the ASQ. Try this on your last inconclusive experiment: write one problem statement and work through at least three whys before you repeat. anything.

Tool 2: Fishbone Diagram

The 5 Whys works well when you have a clear starting point. But what if you are not sure where to begin? Use a Fishbone diagram.

Unlike the 5 Whys, which drills down a single path, a Fishbone diagram helps you map multiple possible causes at once. Place the problem statement at the center and explore possibilities across six categories, known as the 6 Ms:

1. People (Man-power): Lack Training or Communication Gaps.
2. Method: Instructions are unclear, or processes deviate.
3. Machine: Failure of equipment and hardware, for example, a broken incubator.
4. Material: Expired buffers, contaminated controls, or lot-to-lot reagent variation.
5. Measurement: Mistakes in mathematical calculations, statistical analysis, or inappropriate analysis methods.
6. Mother Nature: Humidity, light, blackouts, pandemics, and environmental factors.

Limitations: The fishbone is a divergent tool, meaning it provides possible causes but does not rank or test them. For a complex system with several interrelated variables, this can yield an overly long list of low-probability candidates. Always follow the Fishbone with targeted testing of individual variables, rather than trying to address every variable at once.

Figure 2: Worked example of the Fishbone diagram illustrating the 6 Ms of root cause analysis, adapted from the ASQ. Start by listing one or two possible causes under each category before jumping straight into repeating the experiment.

Tool 3: The 5 Hows

A common gap in troubleshooting is stopping at identifying the root cause without implementing a preventive fix. Finding the cause is essential, but you also need to ensure the same problem does not recur.

That is where the 5 Hows concept comes in. Once you have identified a root cause, keep asking how to implement a practical fix until you reach a concrete, sustainable solution. For example, if the root cause is a buffer left out overnight, then the corrective action begins with proper buffer storage:

• How? Keep reagents on ice or return them immediately per the standard operating procedure (SOP).
• How? Make the SOP accessible by keeping a printed copy at the bench.
• How else? Include a checklist at the lab exit for the last person out each day to confirm reagents are stored correctly.

Whatever the root cause, the goal is to build a step into your workflow that makes the error harder to repeat. Once you have applied the fix, re-run the experiment to confirm the issue is resolved before making any further changes.

Limitation: The 5 Hows can lead to impractical or over-engineered solutions when applied without resource constraints. Each corrective action should be evaluated for feasibility, cost, and proportionality to the risk before implementation.

Figure 3: Worked example of The 5 Hows adapted from the ASQ.

Further reading

• American Society for Quality. 2026. Fishbone diagram (cause and effect diagram). American Society for Quality. Available at: https://asq.org/quality-resources/fishbone?srsltid=AfmBOopxV_T-NRtHDA6lixs-g-L_LKsQYcafknP6YzvU7Is0luM
• American Society for Quality. 2026. Five whys: finding the root cause. American Society for Quality. Available at: https://asq.org/quality-resources/five-whys?srsltid=AfmBOoqk_4o6nEmoWU3zfdzeaqAnO-ldc_AttXhiaFuFSca
• Chesher D. Evaluating assay precision. Clin Biochem Rev. 2008 Aug;29 Suppl 1(Suppl 1):S23-6. PMID: 18852851; PMCID: PMC2556577.
• Clinical and Laboratory Standards Institute (CLSI). 2014. Evaluation of Precision of Quantitative Measurement Procedures; Approved Guideline—Third Edition. CLSI document EP05-A3. Clinical and Laboratory Standards Institute.
• National Library of Medicine. Finding and Using Health Statistics. National Library of Medicine (US). Available from: https://www.nlm.nih.gov/nichsr/stats_tutorial/index.html
• Rosner, B. 2011. Fundamentals of biostatistics (7th ed.). Brooks/Cole, Cengage Learning.

References

Chesher D. 2008. Evaluating assay precision. Clinical Biochemistry Reviews. 29(Suppl 1):S23–S26.

Monaghan TF Rahman SN Agudelo CW Wein AJ Lazar JM Everaert K Dmochowski RR. 2021. Foundational statistical principles in medical research: sensitivity, specificity, positive predictive value, and negative predictive value. Medicina. 57(5):503.

Plebani M. 2015. Diagnostic errors and laboratory medicine – causes and strategies. eJIFCC. 26(1):7–14.

Priya Halvorsen

Priya is a North Carolina-based Technology Transfer Scientist with an NCSU PhD. Using her expertise in microbiology, immunology, and molecular biology, she leads assay development and tech transfer projects in diagnostics, translating complex scientific research into commercialized technologies.

Share this article: