As winter starts to melt away, many of us begin to realize that swimsuit season is right around the corner. But when it comes to watching your weight in the survey world, we have a far easier solution than hitting the gym!
We also weigh our decisions, values, and relationships. But how does this idea relate to your survey? And why should you be watching your weight?
The answer is actually quite simple. The idea behind using weights is that you can assign a numerical value to answer choices, which will then be reflected in your analysis. Here’s what to consider when using weights.
Why bother weighting survey responses?
Percentages, the most common way to analyze survey data, represent a portion of the respondents—but weights allow you to examine your data in a way that includes the responses of every single person who answered a question.
This can provide you with a different perspective or new insight that you wouldn’t have gleaned if you just looked at the percentages or raw numbers.
Weights relate specifically to our Matrix and Ranking question types. With matrix-style questions, you are able to create your own custom weight for each answer option in the scale. And with a ranking question, a fixed weight is already assigned for each answer option.
Still intrigued? Of course you are! Now let’s break down how to weight responses for each of these question types.
How to weight responses: Matrix style
The Matrix (also known as the Rating Scale) question type allows you to create a question with the same rating scale applied to multiple questions. Each question is a row in the matrix, and the scale is formed by each of the columns. This question type will allow you to customize the value of the weight you attribute to each answer option.
As seen in the screenshot above, the values are assigned as 1-5, though you’ll notice you are able to edit this value. This is helpful for a couple of reasons.
Oftentimes, you may want to create a scale that attributes both negative and positive values to answer choices, such as creating weights between -2 through 2. This allows you to attribute a negative or positive feeling to a set of choices.
For example, if you asked your employees if they were content with management and wanted to be sure to pick up on employee dissatisfaction, you could assign a larger negative value to the negative feeling.
Before you assign a weight to your Matrix question, make sure you think through whether you even need a weight. For example, if you ask a Matrix question about which cell phone companies you associate with words like high quality or low cost, getting a weighted average of 3.4 won’t tell you whether people associate quality with Apple or Samsung.
One last thing to note when using the Matrix question—you’re able to add, adjust, or remove weights from the design section of your survey while the survey is live.
Weighting survey responses for your Ranking questions
When it comes to the Ranking question type, the weights have already been applied to the numbered scale for you. These questions typically ask respondents to rank a list of options based on the item they prefer most to the item they prefer the least.
In a Ranking question, the more an individual prefers an option (i.e., the closer the option is to 1) the larger the weight associated with this selection. Think about it as the weight of importance: The more important the option is to a respondent, the more weight is attributed to it (and vice versa).
So what are some of the reasons to use a Ranking question over a Matrix question, you ask? Well, a Ranking question is more straightforward than a Rating question—it simply looks at the respondent’s preference between a group of options. Because the respondents have the same comparisons or group of options in mind, often this can result in more reliable data.
Also, you don’t need to assign a weight to each answer choice, as this is automatically already done for you. But note, while it is a no brainer for the survey creator to assign weights in a Ranking question, it may be difficult for the respondent to answer, particularly if two options are equally important since each option needs to be assigned a different rank.