Overview: Competitive Topography Research Test

Competitive Topography is another turnkey solution created to explore a number of entities, such as brands, that can be rated on a list of attributes. This methodology will give you an important understanding of consumers' perceptions of each brand, perceived similarities between them, each brand's most differentiating attributes, and a visual comparison of ratings among the restaurants in aggregate as well as individually by each attribute.

To accomplish this deep analysis, researchers used to run complicated, expensive, and time consuming data modeling projects. At times these projects needed to be re-run if some of the parameters changed during the experiment. One of the best advantages of aytm Competitive Topography test is that it's incredibly easy to set up. It can be added just like any of our other question types - simply by dragging and dropping the corresponding icon from the sidebar, or by adding it at the bottom of the survey. An existing question can also be converted into a competitive topography question. All you have to do now is to fill out the list of brands you want to test and the list of attributes. If you're not sure which items to add, or you have a very long list, it may be wise to first run a MaxDiff survey to narrow down a longer list of attributes to just 5 to 7 of the most important ones. Alternatively, you can ask an unaided open-ended question, and code the answers, arriving at the list of the most frequently-referenced brands and attributes. If you already know what you want to test, you can proceed without this extra step.

The model works best with at least 4 entities and 4 attributes, and you can add up to 10 items in each list. Please note that we recommend having ~7 items in each list in order to get the most informative visualizations. In the right bottom corner of the attributes list you will find a combobox with a preset library of common attributes broken down to three groups - product attributes, service attributes, and general brand attributes.

As with most other question types, you have the ability to illustrate every field and randomize the order in which they will be presented. If you don't want certain options to be randomized, you can easily anchor them in place so they appear in the same order for each respondent.

 

You'll have to choose whether you want respondents to rate each brand with stars (which is the default mode on our platform) or sliders. If you choose to use a star rating, you'll be able to use 5, 7 or 9 stars for each attribute. Choosing sliders will give you more flexibility. You'll be able to choose from our library of pre-written Likert scales, edit the answers, write your own custom answers, and even adjust the scores from 1 to 99, which are provided automatically for you when an answer is chosen. By default we assume 10 points for one star OR the lowest answer on your Likert Scale, and we go all the way up to 50, 70, or 90 for a top rating. If you decide to edit the Likert scale OR adjust the scoring, please make sure you know exactly what you're doing, since it may drastically affect your model and data visualization.

The last important decision here is to choose how to group these two lists. By default, we'll group them by entities - or brands, in our example. That means that each brand will be presented as a separate question, with attributes listed as sub-questions below. This grouping may be easier on respondents since it helps them to activate memories of all their experiences with a given brand or entity, enabling respondents to rate each brand or entity by all attributes you're testing.

If you switch to 'group by attribute', each question will ask about one attribute at a time, such as "Food healthiness", and will contain all compared brands on the page. It might introduce a higher cognitive toll on respondents, since they'll have to access a lot more memories across brands in order to answer each of the questions. Even so, in some cases this might be more valuable, since it'll help focus attention on comparing all brands across a given attribute. Please note that this experiment will take as many questions as there are items in the list you're presenting.

It's important to make sure the question is still appropriate for your case. We have four pre-written question texts, designed for star ratings and sliders in both modes - grouped by entities and by attributes. Our platform will suggest default text as you adjust the parameters of the experiment, as long as the field is empty or untouched. If you edited the field already, we won't mess with your text, but you'll have to understand how it works and carefully test it out. You may notice the internal piping here. If you group by entities, the word [entity] will be replaced by a brand as you roll over it, helping you preview how each of the questions will read in the survey. The same thing will happen with your attributes. If you accidentally remove the magic word in square brackets, don't panic. You can type it back in or click on the warning that will appear underneath. You can click on the question icon to read a quick blurb about how it works. Clicking on the text of the warning will insert the code at the end of your question. Make sure you move it to the appropriate part of your sentence. We recommend watching the tutorial video for this question type to see it in action.

To get the most reliable insights, we recommend that you run this research model with ~750 completes.

As data starts streaming in, we'll crunch the numbers in a multidimensional space to negotiate all distances between entities and attributes and will plot a 3D data visualizations to help you explore and present the findings.

There are three distinct ways you can visualize the results: Multidimensional Scale, Topography View, and Quadrant Analysis. You can toggle among them at the top of this test.

The multi-dimensional scale is closest to classic perceptual mapping visualization. Essentially the rendering treats each of the attributes you added as a separate dimension or axis in a hard-to-imagine multidimensional space. It looks for a balance among all the forces and puts your tested entities into very specific positions in relation to each other within this space. Then our rendering runs up to 200 attempts to find the most accurate two-dimensional version it can create of this multidimensional model, to make it comprehensible. We list the R square (or 'projection accuracy') in percents. It usually varies from 95 to 100 percent.

Let's take a closer look at the visualization. The first thing you'll notice is that the brands are depicted with color dots and have horizontal labels. Attributes are shown as axes, meeting in the center and labeled along each line. Some attributes might be so close to each other in the minds of survey takers that we'll have to blend them into a single line and list both labels one after another. In such cases, the labels will go in the same order as the blue dots on the axis.

Please note that some axes depicting attributes are in bold - these were the most differentiated attributes according to respondents. We can't automatically assume that these are the most important to the industry, or consumers - let's just say these had the most variation from entity to entity (brand to brand), and therefore we're very confident that if an entity/brand performs well or poorly on a boldface trait, it stands out considerably from the others in respondents' minds.

Next you may notice that some axes are shorter than others. Each of them ends with a blue dot, which we call the 'epicenter' of an attribute. If an attribute is short and close to the center, it means this attribute was similar among all your tested brands and wasn't associated strongly with any particular one.

When analyzing this visualization, please pay close attention to the proximity of the brands and attributes' EPICENTERS, not the axes. For example, on the illustration above, Taco Bell is equally close to both the 'Convenience/location' and 'Price' axes, but it's much closer to the epicenter of the 'Prices' attribute, which is what matters. We can interpret it as that Taco Bell was closely associated with good pricing. We can also conclude that in consumers' perceptions both Burger King and Taco Bell were seen as similar to each other, with strong associations of affordability, variety, speed, and convenience of locations. They were not specifically associated with 'food quality/taste' or 'food healthiness,' which are on the opposite side of the map. Subway, on the other hand, was different from all the other brands tested here. We can see that it came closest to the epicenter of food healthiness, which was one of the strongest differentiating attributes in our study. Arby's and Chick-fil-A were most closely affiliated with taste and cleanliness, and McDonald's with convenience of location and prices. In fact you can read the most associated attributes for each brand from the table below. We'll list them in descending order, with the average score each brand received for the corresponding attribute.

Next thing that you'll notice is the color. Subway was rated higher than other brands by most of the attributes, which is why we present it in green. Please refer to the legend for the relative breakdown.

If we click on McDonald's, we'll see a very different picture: while it was in the top percentiles for 'Convenience/location' and 'Prices,' it was rated lower compared to other brands in 'Food healthiness' and 'Taste.

 

This is something that we invented here at aytm. It's based on the same multidimensional scale model, but presented in an interactive, intuitive 3D model. By adding a third dimension we enable you to visualize the scores as the heights of a landscape while keeping the proximities intact between brands and attributes. The entities or brands are presented with color pins and horizontal labels, while the epicenters of attributes are marked with white flags and labeled vertically. You can rotate the model and explore it from any angle. You can zoom in or out by using the scroll wheel of your mouse or scroll gesture on a trackpad. The color helps highlight different heights to better represent mean scores, and you can refer to the legend on the bottom for the exact range of values. The higher the mountain and closer to the red side of the spectrum, the higher the mean score an entity has received. The deeper the valley and the closer to dark blue it is, the lower the mean score. From this summary view we can easily assess that Subway was far ahead of the other brands by total score, since it resides on the top of a high mountain while the rest are on the foothills or in the valley. You may also see that the closest attribute to the Subway mountain was again "Food healthiness," while McDonald's was very close to the epicenter of the "Convenience/location" attribute.

You can look at your insights from the perspective of either brands or attributes, and it doesn't matter how you grouped your lists during survey collection. Entities or brands is the default setting. The combobox on the right allows you to control what variable we will visualize using the height of the terrain.

Switching from the total mean scores to any of the individual brands will show a different picture every time. For example, here's the terrain for McDonald's. It has a peak of Convenience of location, indicating that this was perceived as McDonald's strongest trait, and a valley of food healthiness, indicating that this was perceived as McDonald's weakest trait. All the other qualities are arranged on different heights in between. Please note that in the table below you'll see mean scores for each of the brand's attributes.

As before, you can select any individual attribute from the list on the right and see which brand dominated that attribute. We already learned that Food healthiness is strongly associated with Subway, followed by Chick-fil-A and Wendy's. When we check out Prices, on the other hand, we can see which brands are perceived as most and least affordable. Even though Subway has a lower rating for prices in comparison to its ratings on other attributes, when compared to the other brands we tested, its prices are still rated very well. Check the exact scores in the table below to get precise readings.

Two combo boxes below allow you to declutter the visualization, which is especially useful for the summary view modes when too many elements are competing for your attention. Hiding an entity or attribute label will not alter the underlying model, it'll simply remove the element from the screen so that you can export exactly what you need to illustrate and communicate the finding in your presentation.

You can export any view as an image, or as a vector graphic which can be scaled to any size without compromising the quality. Please note that when you export the survey into PowerPoint, only the current 3D view will be included. If you'd like to show more than one view, you'll need to manually export each image and paste it into your presentation.

It wouldn't be fun if we hadn't added an extra twist to this view mode, on top of what you'd normally expect from a quadrant analysis. When you click on a circle, you can see the actual distribution of answer combinations by the current x & y attributes. Why does it matter? Since the position of the brands is a mean of all collected ratings, sometimes it's unclear what distribution of answers resulted in such a mean. For example, respondents could have been very polarized in their ratings, or they all consistently gave an average rating. In both cases the mean score and location of the brand would be very similar, and you wouldn't know the underlying truth. To discover the exact number of people who gave a certain combination of ratings, hover over the grey dots on the grid. The circle in the left bottom corner, for example, represents one star for "Convenience" and one star for "food healthiness," and 36 people gave that combination of ratings. The largest grey circle seems to be in the center of the top row, and this is because it was the most popular combination - 148 people gave the chain 7 stars for "convenience" and 4 for "healthiness."

Another useful thing about quadrant view is that you can see mean ratings for up to 4 selected attributes at a time in the grid below. Click on the headers to sort in ascending or descending order.

Probably the most amazing thing about Competitive Topography and other research tests that you can run on the aytm platform, besides being great interactive visualizations and easy to use, is that they're fully integrated into the stats page. This means that you can apply any combination of filters by demographics and/or traits, and have the numbers re-crunched in almost real time for you. You can see, for example, how perception of these fast food restaurants varied between genders or among age groups. You could even select a subset of respondents, such as those who are most loyal to Chick-fil-A, and view their overall perspective on our competitive set of quick service restaurants.

We encourage you to play with the demo survey and watch the tutorial video above. Feel free to ping us with questions or to set up a personal demo.