Mastering the Basics of Statistical Analysis

Why Statistics Matter in Consultancy

In consultancy, credibility is currency. And nothing builds credibility faster than data! Statistical analysis gives us a language for process behaviour. It helps us distinguish between what is normal variation and what signals a problem requiring intervention.

As consultants we often work across industries and departments where process complexity can hide root causes. Statistical thinking provides the structure to cut through the noise and guide our clients toward sustainable solutions. It enables:
– Data-driven decision-making
– Evidence-based recommendations
– Quantification of risk and opportunity
– Improved client confidence in findings – this really is essential.

Whether you’re working with transactional data in a service environment or physical measurements on a production line, statistics allow you to base your conclusions and build the business case for change often also highlighting the cost and consequences of not.

Know Your Data: Discrete vs Continuous

Understanding the nature of the data you are dealing with is the first step in selecting the right analysis method and you would not believe how many times in my early career I missed this vital step out.

Discrete data refers to countable items. It is finite and often categorised in whole numbers. Examples include:
– Number of customer complaints
– Defective units in a batch
– Delivery errors per day

Continuous data can take on any value within a range and is measured with precision. Examples include:
– Cycle time in minutes
– Product weight in grams
– Temperature readings in degrees

But why does this matter? Because different types of data require different statistical tools:
– Use p-charts or u-charts for discrete (attribute) data.
– Use X-bar and R charts, or IMR charts for continuous (variable) data.

Choosing the wrong chart type can lead to incorrect conclusions about process stability or capability ultimately negatively impacting credibility, as I have personally experienced!

Sampling and the Rule of 30

Sure, in an ideal world, we would measure an entire population. In reality, time, cost, and practicality require us to sample. But sampling introduces risk: the smaller the sample, the more variability (the enemy) and uncertainty we introduce.  I have seen at Skewb someone doing a 100% “sample”.  Imagine the time that would have been saved if they knew about sampling.

The Rule of 30, as those who have worked with me will have no doubt heard me say is based on the Central Limit Theorem (CLT), which states that as the sample size increases, the distribution of the sample mean becomes approximately normal, even if the underlying population distribution is not. This is especially useful when applying parametric tests, which assume normality.

Using 30 or more samples:
– Reduces sampling error
– Improves reliability of statistical inference
– Allows use of control charts and hypothesis testing

This rule isn’t absolute, but it’s a strong guideline, particularly for initial process analysis though personal experience has at times, made it challenging to advise why the sample size is so “low”.

Sample Size Calculators: Smarter Planning

Accurate conclusions start with good design! A properly sized sample is essential for:
– Detecting meaningful differences
– Avoiding false negatives (Type II error – accepting the null hypothesis incorrectly)
– Controlling study cost by reducing time.

A sample that is too small may lead you to overlook real problems. A sample that is too large wastes time and resources. Sample size calculators help you balance precision and practicality by considering:
– Desired confidence level (typically 95%)
– Acceptable margin of error
– Population variability (standard deviation)
– The type of hypothesis test or chart you plan to use if applicable.

Many online tools and statistical software packages like Minitab and Excel include built-in calculators. In consultancy, using these tools early can demonstrate professionalism to clients and ensure project timelines remain realistic.  I have never been challenged when quoting a confidence interval of 95% so this would be my recommendation for you.

Trust Your Measurements: Gauge R&R

The most sophisticated analysis is meaningless if your measurement system is flawed. Before you assess process performance, you must first assess how trustworthy your measurement data is.

Gauge Repeatability and Reproducibility (Gauge R&R) is a key component of Measurement System Analysis (MSA). It evaluates how much variation in your measurements is caused by the measurement system itself.  What I class as a fail, someone else could class as a pass.

– Repeatability: Variation when the same operator measures the same part multiple times using the same equipment.
– Reproducibility: Variation when different operators measure the same part using the same equipment.

Gauge R&R studies are particularly important when:
– Introducing new gauges or measurement tools
– Training or assessing operators
– Establishing baseline process capability though this is not capability analysis

A well-executed Gauge R&R study involves selecting a representative sample of parts multiple operators, and multiple trials. Results are typically broken down into percentage of total variation:
– <10%: Acceptable
– 10-30%: May be acceptable based on application
– >30%: Not acceptable; improvement required

IMR Charts: Monitoring One Step at a Time

Individual and Moving Range (IMR) charts (my personal favorite) are a type of control chart used when subgrouping is not possible or practical.

The I chart displays individual measurements over time, while the MR chart shows the range between successive measurements, highlighting variability.

IMR charts are useful when:
– Measurements are costly or infrequent
– You need quick feedback on process changes

These charts help identify special cause variation and process shifts. For example, a point outside control limits, or a run of 7 consecutive points in one direction (above or below the mean), would indicate a non-random change worthy of investigation.  Some of you will no my dislike for the far too often used term “trend” and this is where it originates!

IMR charts are particularly valuable in service-based or transactional environments where data is collected continuously rather than in batches.  The only downside is that they look like they have been produced on a computer from the 1980’s!

The Consultant’s Statistical Toolkit

Every consultant carrying out statistical analysis would do well to be equipped with:
– Excel & Minitab: For statistical tests, control charts, and capability analysis
– Sample size calculators: Online or integrated in software
– Reference texts: Donald Wheeler’s Understanding Variation (I cannot recommend this book enough)
– Standardised templates: For SPC, MSA, and data collection

These tools increase your efficiency and signal credibility to our clients.

Final Thoughts

Statistical analysis isn’t just for data scientists. It’s a vital skill that strengthens recommendations, reveals root causes and variation, and builds lasting value for clients. With a strong foundation in statistics, you will become not just an advisor, but a trusted partner in transformation.

By Tim Pritchard

Tim Pritchard is Principal Consultant at Skewb. To speak with him about this topic in more detail, you can connect with him on LinkedIn.