Hypothesis testing is the study and analysis of assumptions. Before executing an action or believing a claim, you test your theory by examining the basic rules and framework of a claim.
Hypothesis testing is categorised into two basic types:
- Null Hypothesis
- Alternative Hypothesis
Null Hypothesis has well-defined parameters and is said to be the default-claim.
For example, you want to find out if your team is training properly by logging in the time they spend in the race track. Your expected default training duration for the team is 80 minutes, as reported by the captain of the team. This is your null hypothesis. An Alternative Hypothesis is the opposite of your null hypothesis. It basically invalidates or disapproves the null hypothesis by providing evidence related to the new claim.
When proven true, it rejects and replaces the Null Hypothesis.
For example, if you hire a new maintenance worker for your candy manufacturing company and after a week’s inspection, he reports that the candy machine no longer makes 10g chocolate every day but rather 4g chocolate bars instead, then that’s an Alternative Hypothesis.
Hypothesis testing is used to make design decisions every day and extends even to the supply chain level and beyond. Data storage and collection are considered not enough to make business decisions. In summary, Hypothesis testing is a way of evaluating and analysing collected data samples in order to make business decisions.
The key objectives of hypothesis testing are:
- To identify and name the elements used in the hypothesis
- To differentiate between the null hypothesis and the alternative hypothesis
- Lay out the steps for hypothesis testing
- Identify common errors and pitfalls in these tests
- Apply the results of the hypothesis testing to solve common design, business, or organisational problems
For instance, if a company wants to boost their sales by 30% nationwide, they would launch a new campaign in a new region and see how it performs. If the sales profit in that region meets the 30% threshold or surpasses it, then the company would collect the data, analyse it, and launch new campaigns nationwide provided it’s convinced that the hypothesis testing results yield of a profit of 30%+, with the given data and samples.
It is a measure collected from sample data which is used to decide how to create a comparison between the null hypothesis and alternative hypothesis. For example, if you believe the candy bar manufacturing machine is malfunctioning, your test statistic would be sampling 50 bars to check if the average weight of these bars is really 10g or 4g as stated by your old and new employees, respectively.
This is a term that draws a baseline value which concludes whether the null hypothesis or alternative hypothesis is correct. After weighing those 50 candy bars through the week, if you find that the average weight arrived at 4.25g, then that number is statistically significant and evidence enough to disprove the null hypothesis.
Level of Significance
The level of significance is represented by the symbol Alpha (A). It is calculated by subtracting the number 1 from the confidence interval which is represented in decimals.
For instance, if C = 95% or 0.95
1 – 0.95 = 0.05
Level of Significance and Confidence Intervals are together used to cement the hypothesis approval and substantiate the boundaries which were dubbed Statistically Significant.
Hypothesis testing is most of the time tied to data and evidence. You reduce a lot of paperback and decision making too by involving your team in the process. From journey flows, empathy mapping, analysis of customer survey data, and more, this testing method lets you evaluate if the proposed UX Design works and functions the way it’s expected. Of course, the context makes the parameters of these tests change, and that’s important to note.