The index tells you how likely it is that members of your audience will exhibit a particular trait compared to the base. Here’s what you need to know about it and some common FAQs.
- When's an index meaningful?
- Why are the indexes for my audience all so flat?
- Why are the indexes for my audience all so high?
- Why are the indexes for my audience all exactly the same?
- Can I calculate indexes across audiences?
- Why are my manual calculations giving me different results?
It starts at 100
An index above 100 means that members of your audience are more likely to exhibit a particular trait while an index below 100 means they’re less likely to do so.
The indexes in the below crosstab tell us that Gen Zs are 8% less likely to buy new tech products as soon as they’re available while Millennials are 22% more likely to do so. Note we're rounding the decimals here.
It’s a propensity, not a proportion
The index refers to your audience as a whole, so make sure you're quoting it correctly.
✔️ Correct: “Millennials are 22% more likely to buy new tech products as soon as they become available”
❌ Incorrect: “22% of Millennials are more likely to buy new tech products as soon as they become available”
It can be read, and calculated, both ways
As well as telling you how likely it is that members of your audience will exhibit a particular trait, the index also tells you how likely it is that people with that trait will be in your audience. This is because the index reflects the strength of the relationship between two groups, and relationships are a two way thing - the math shows it.
To calculate the index, you can either:
- Divide the column % of the crossover cell by the column % of the total cell for the relevant data point, and then multiply by 100
- Divide the row % of the crossover cell by the row % of the total cell for your audience, and then multiply by 100
Let’s return to our example…
Using the results from the above crosstab, you can divide 20.5 with 16.9 and multiply by 100 to get the index for Millennials and the shown attitude statement. Alternatively, you can divide 44.3 with 36.4 and multiply by 100 and get the same result. This means you can read the index both ways.
Note: Your answer may be a little different because the platform uses unrounded versions of these figures for more accurate results.
✔️ Correct: “Millennials are 22% more likely to buy tech products as soon as they become available”
✔️ Also correct: “Those who buy products as soon as they become available are 22% more likely to be Millennials”
When’s an index meaningful?
It depends on how common the trait you’re looking at actually is. If only 1% of the base have a particular trait, but 2% of your audience do, you’ll have an index of 200. That may seem big, but it doesn’t change the fact that only 2% of your audience actually have that trait.
On the other hand, if 50% of the base have a particular trait, but 60% of your audience do, you’ll have an index of 120. That may not seem as exciting as an index of 200, but we’re talking about a 10 percentage point difference compared to the average and over half of your audience. All in all, that’s pretty big.
The index is helpful at showing how your audience differs from the average but, for the full picture, you need to consider the other metrics as well. As a rule of thumb however, you won’t usually need to pay much attention to any indexes between 90 and 110.
Why are the indexes for my audience all so flat?
If the indexes for your audience are all quite flat (i.e. close to 100), it could be because you’ve created a very broad audience. Consider making your audience more targeted by adding in more data points, or breaking it down into two or more sub audiences.
Alternatively, if you’re looking at your audience at a global level, it could be that regional differences in your audience are cancelling each other out. Try looking at a specific region or market instead.
Why are the indexes for my audience all so high?
Your audience could simply be very engaged with the category you’re comparing it to.
This is often the case when looking at a relatively young audience and comparing it to online behaviors. In such a case, your audience may not be any more likely to engage with those behaviors than other young people, but the fact they’re young means they’re more likely to do so than the rest of the online population, leading you to see consistently high indexes.
To gain a different perspective, you could apply a base of young people to your analysis to see what makes your audience different from its peers.
Why are the indexes for my audience all exactly the same?
This means your audience includes everyone who responded positively to one of the options in the question you’re looking at, and nobody else. Let’s look at an example…
If you create an audience of people who are employed and compare it to work sector / industry, the index will be the same for all options (in this case, 138.8). This is because the work sector / industry question is only asked to those who are employed.
Think about it like this: Your audience is only more likely to work in accounting because they work - not because they have a particularly unique relationship with accounting. They’re not any more likely to work in accounting than any other industry.
In these instances, consider applying your audience as a base instead.
Can I calculate indexes across audiences?
Want to know how likely one audience is to exhibit a particular trait compared to another? No problem! Just take the calculation outlined above, but instead of dividing by the corresponding figure from the totals column/row, divide by the corresponding figure from the other audience.
For example, if we divide 44.3 by 22.6 and multiply by 100 we get 169. We can therefore say that “Millennials are 69% more likely to buy new tech products as soon as they become available than Gen Zs”.
Why are my manual calculations giving me different results?
This usually happens when you compare a question that’s been asked of all respondents with one that’s only been asked of a representative subsample. For example, you could be comparing a question from the main Core survey with a question from the Brand & Media module, or something from an add-on with its respective “primary” data set (e.g. Core Plus vs Core, or USA Plus vs USA).
Note: If you’ve not come across these terms before, we’d recommend reading this article first!
As far as indexes are concerned, this means that respondents outside of the relevant subsample have to be excluded from the calculation. Our platform does this automatically, so it’s not something you need to worry about. However, it does mean that manual calculations won’t correspond with what’s being shown in the platform given these adjustments all take place behind the scenes.
If you want to take a different approach - e.g. you want to be able to make manual calculations or need 100% consistency throughout your entire analysis - you can apply the relevant subsample as a base using the corresponding “audience size” data point from the “survey details” folder. This would allow you to apply a base of Brand & Media module or Core Plus respondents to your whole analysis, for instance.
However, it's important to note you don’t have to do this as our platform automatically uses the largest available sample for each calculation individually. In other words, when comparing questions asked of different subsamples, our platform automatically applies the smaller of those two subsamples as a base to the relevant cells even if that base isn’t applied to the analysis as a whole. For more detail on all things rebasing, check out this article.