---

What Is 'Sampling'? (Measurement)

Sampling is the process of taking a small percentage of the total output and using it as a proxy for the entire system. Sampling can help you identify systemic errors quickly. Sampling is good for quick tests of quality without incurring huge costs.

Always make sure to test a random and uniform sample.

Josh Kaufman Explains 'Sampling'

What do you do if your system is too large or complex to collect data on every process? Sometimes it's not practical to measure the Flows of an entire system. If you're managing a Scalable system, it's often not possible to test every unit or find every error.

How can you identify potential problems quickly when your system is cranking out millions of units or completing millions of transactions?

Sampling is the process of taking a small random percentage of the total output, then using it as a proxy for the entire system.

If you've ever had your blood taken at the doctor, you'll have a good idea of what sampling entails. The doctor draws a small amount of blood, which is then sent to a laboratory for analysis. If testing reveals anomalies in the sample, it's likely that the same issue is present in the bloodstream in general.

Sampling can help you identify systemic errors quickly without testing all of the output of the system, which can be time-consuming and expensive.

If you're manufacturing mobile phones, you don't necessarily have to test every single phone that comes off the line. If you test one in twenty phones, you can still identify errors quickly enough to fix the system if something goes wrong. Depending on how quickly and accurately you need to spot errors, you can increase or decrease the sampling rate.

Random "spot checks" are also a form of sampling. Many retail stores employ "secret shoppers" to periodically test customer service or the skills of their sales staff. These shoppers are typically hired to express interest in specific items, ask certain questions, make a return, or generally be annoying.

Since the staff doesn't know which customers are real and which ones are tests, it's an effective way for the management of a store to check for quality without scrutinizing the staff every second of every day.

Sampling is prone to bias if the sample is not truly random or uniform.

For example, measuring the average household income in the U.S. by surveying property owners in Manhattan will skew the data considerably higher than if you survey property owners in West Virginia.

For best results, use the largest random sample you possibly can. If you need to test for quality, sampling can help you collect good data without incurring enormous costs, provided you stay on the lookout for potential bias that can skew the data.

Questions About 'Sampling'

• Can you test the total output of your system, or would a sample be sufficient?
• What is the largest sample you can feasibly collect?

---

"If you don't believe in sampling theory, next time you go to the doctor and he wants to take a little blood, tell him to take it all."

Gian Fulgoni, founder and chairman of comScore, Inc.

---

---

Share this concept:

https://personalmba.com/sampling/

---

---

---

---