Interpreting Index & Composition

The best practice for interpreting Index and Composition is to consider both metrics and weigh their significance according to your business needs, use case and the insight you’re trying to understand.  

Index can help you find the needle in the haystack. It helps you to identify attribute values that stand out from others and make your audience unique compared to your baseline. Focusing on attributes that are unique to your audience will ensure that your targeting is precise and efficient. This strategy can help you find ways to break through the clutter and reach your desired audience. You should give more weight to an attribute’s Index when you are aiming for precision and efficiency. 

Index is a more telling metric when analyzing attributes that: 

1. Are multi-select survey questions (meaning that the attributes have numerous values from which audiences can choose), 
2. are a wide-spread trait for the majority of the population.   

Usually, these attributes are retail brands, stores, restaurants, hotels or media consumption traits, such as magazines or TV shows. 

Composition tells you what percentage of your audience has the attribute that you’re analyzing. It helps you determine how large of a net you’d be casting if you were to activate against an audience based on one of its attributes. It is your go-to metric when the aim of your research, or campaign, is scale.   

Pro-tips: 

1. Once you have identified the Indexes worth considering in your analysis, take a look at the Composition to verify that the insight is valuable to your research.  
2. You can use the filter function in the Analytics tab within Intelligence Center to change the order of your insights and place the most relevant ones on the top of your list. 
3. Finding meaningfulness in the Index metric varies based on data type and business case. For example, as a rule 120 and above is considered a meaningful over-index value for survey data because there is more differentiation between survey values, while 105 can be meaningful for behavioral or tag data. 
4. The range of Index values is also important. If there is a broad range, like 80-305, then an Index of 105 is not meaningful because there are much higher index values. If the range is between 95 and 105, then an index of 100 probably becomes meaningful. 

These are general guidelines, but finding meaningfulness in Index and Composition ultimately depends on your business context. Below we've included three examples to demonstrate how to balance Index and Composition when interpreting your insights. 

Example 1  

Scenario: You want to decide on which social media channel to run your ad campaign. You start by looking at your audience’s social media membership and identifying the attribute value with the largest Composition. The higher the percentage is, the larger portion of your target audience you will reach with your ad. If scale or reach is your main goal, this might be your strategy.   

However, if we only rely on Composition here, we'd report back to our boss that we should run our social budget on Facebook...which is probably something they already know! If we also look at Index, we can see that our audience is 21% more likely to use Pinterest, and this attribute value also has a healthy Composition. This tells us that, if we're not already doing so, we should incorporate Pinterest into our social media planning and budget.  

Example 2   

Scenario: We want to advertise to our target audience on a streaming service. Our analysis reveals that 20% of our audience uses Spotify and 19% uses Amazon Prime to download or stream music. While these Composition numbers are solid, the Indexes are low, hence these attributes are not that unique to our audience compared to the baseline.  

The audience is over-indexing for "Tidal" in the image above. However, its Composition is <1%. This tells us that less than 1% of our audience actually uses this streaming service, so our marketing efforts would not be as efficient on this channel. 

On the other hand, our audience is indexing at 128 for SoundCloud, with a healthy composition of 7%, which makes this attribute value more unique to them. So, we recommend that if you’ve been allocating your streaming budget to Spotify and Amazon Prime Music, you should now consider SoundCloud as well. 

Example 3  

Looking at an audience’s TV show consumption habits, filtered by Index (descending), we will see that the Composition is a bit low for the top indexing shows, but the Index is remarkably high. However, a Composition of 5% is good for attributes like “TV Shows Watch Regularly” because there are so many shows an audience can choose from. In this example, we will rely more heavily on Index to focus on the attribute values that make our audience unique compared to the baseline. 

Index and Composition in Curated Reports 

In most cases, components in our curated reports are shown by Index, with the minimum Composition threshold also indicated nearby. You will see that the minimum threshold can vary between 3% to 15%, depending on the report component. When developing each component in our curated reports, our product management team and professional services team work together to determine the appropriate Composition threshold for each report component to ensure that the data you see is both significant and unique to the audience. 

When the minimum Composition is not defined, it’s because all attribute values are listed, and we don't need to rely on the Composition to filter the displayed attributes.  

Sometimes, certain report components will display insights by Composition. These include demographic traits and other attributes where we bucket values into categories, expressing what percentage of the whole an attribute value is, like hours spent online per week. 

When interpreting data that displays both Index and Composition, you should look for a balance between the two metrics, bearing in mind the specific use case and insight you’re analyzing. As a best practice, look at the Index to bubble up the differences in attribute values, then examine the Composition for a closer look. As a reminder, if the focus of your research is scale, pay attention to Composition. On the other hand, if it’s precision and efficiency you’re looking for, rely on Index.

Index and Composition in Audiences with Survey and Connected Profile Data

When combining audiences derived from Survey data and Connected Profile data, sometimes we see differences in index values and composition percentages across these attributes. To better understand the underlying differences, let's review our two types of data universes:

1. US Consumer Study Data (Resonate Elements): This is data obtained directly from surveys and is used to project metrics about a larger population to help marketers understand consumer trends.
2. Connected Profiles Data: This data is modeled from online behaviors and device usage observed in our behavioral data feed, leading to more informed guesses about consumer traits and habits. These insights are then used to predict patterns across devices observed online, offering real-time insights without being projected onto a broader population

Survey data & Connected Profiles represent two distinct data sets with their own unique baselines. The survey data employs baselines derived from survey waves specific to your audience and projected to a total US Population. In contrast, Connected Profile data leverages a baseline that is comprised of the total number of observed device IDs available at the time the data was loaded (or refreshed) within our behavioral data feed.

Audiences derived from Connected Profile and Survey data can be tied to slightly different baselines. Although we construct our Connected Profiles to mirror the weighted distributions found in survey data, aiming to closely align with it, they remain independent of each other and there may be slight differences between baselines associated with data from different points. As a result, there may be cases where the indexes display different relationships to the % Compositions due to the distinct nature of the underlying baselines.

In these situations, we recommend that customers use the index as a consistent metric for comparing different audiences. The index reflects the likelihood of an audience possessing a trait relative to its corresponding population.