I'm going to be showing you everything in Displayr, but all the steps can also be done in Q, with one small exception, which I'll point out. But the principles are applicable regardless of which software you use.

 

What is U&A Research

A quick Google search reveals there are lots of definitions of U&A research out there. To cater for many of them, I'll define it as:

• Usually involving a large scale survey, both in length and number of respondents
• Containing generalized measures of the category in question and the brands in it
• Analysis to understand relationships between knowledge or attitude and targeted behaviors
• With the aim of reviewing and informing strategy

 

U&A Research Context

For focus today we are going to concentrate on those general measures, and touch only briefly on the topics of segmentation, driver analysis, and mapping. Our current webinar series has detailed content in these other areas - you can find this our website and I'll also leave you with the links as the end.

 

The U&A Research Challenge

It's fair say that U&A Research can be challenging, for a number of reasons …

Why U&A Research Can Be Daunting

Unlike, for example, ad testing or new product development, the U&A brief can lack focus, and may simply ask to … "tell us everything we need to know about our market"

U&A's can be expensive, and are usually triggered by an event such as a decline in sales or a new CMO arriving, there is a lot of expectation

And many measures multiplied by a many respondents means a large volume of data to wade through

It can be hard to know where or how to begin

 

4 Key Frameworks

Now in a short webinar we are not going to prepare a full U&A report. However, we recommend working through these 4 key analysis frameworks, to get everything off to a good start and give you the confidence to move forward.

We'll work through each framework and see how they build the foundations for a strategic outcome …

 

Competitive Market Structure

With Competitive Market Structure, we want to know who we most compete with, which tells us the true market we are in, and ideally the nature of that competition.

The data requirements to work this out are typically in a matrix of brands by brands or brands by attributes. We prefer attributes, as this will tell us something about the how the brands compete.

From there we map the market, and enhance the map where we can, to guide us

Then locate our brand, our direct and indirect competitors, and in each case determine what they have in common

 

Data requirement: Brand Attribute Grid with brands in the rows

Here is the brand attribute grid we will use …

And we want to have the brands in the rows

 

Data requirement: Market Share + Change in Share

We also want tables showing the market share of each brand and the change in market share over time so that we can use these dynamics in our decision making

 

Market map

Let's now create a market map and explore the competitive market structure

We'll use correspondence analysis. I'll go through some of the technical steps quickly today, but this was the topic of a whole webinar recently, so there is a resource for the keen student

Anything > Advanced Analysis > Dimension Reduction > Correspondence Analysis of a Table

We hook it up to our table of D1 Brand Attributes

Chart > Titles and Labels font size: 10

Grid Lines Off

Next step is to set the row normalization to row principle scaled, which ensures that the map most accurately shows the relationships between the brands

Normalization: Row Principal (Scaled)

Now let’s add market share as bubbles

Output: Bubble chart

Sizes: market.share

And let's color the bubbles to show the change in share over time, with red marking decline and blue growth

Bubble colors: market.share.change

Set midpoint to : Zero

You can see that some brands are in blue. It will be helpful to have them at the top of the map, indicating growth

This is a fake data set with brands disguised as fictitious fast food brands. We are going to assume we are working on behalf of Burger Chef. So, we have a small problem, as the brand we care about is hard to see.

We can rotate the correspondence analysis to maximize it's explanation of Burger Chef, it's easy to do.

That's a bit better. We can also experiment by removing an attribute that might be an outlier, to see if that clears things up

We deleted Healthy Food options because it was a bit an outlier. But, we don't want to lose this information, so we will add it as a textbox.

Text Box > click and drag near Bread Basket. 'Healthy food options'. Properties - Reduce font size to 9, set as Green, Ctrl>D (Duplicate), drag to bottom right

OK, let's work out what it is telling us.

Looking on the right, we can see that Burger Chef is mostly in the Convenience space

Relabel text box: CONVENIENCE

Color: Grey
Size: 12

We are also a bit about restaurant quality

Relabel: RESTAURANT QUALITY

Over here there are brands that compete more on food taste and quality

Relabel: FOOD TASTE AND QUALITY

Down here we have brands competing on value

Relabel: VALUE

So, to summarize

Burger Chef's "Tier 1" competitors are Arnold's and Southern Fried Chicken. They are most strongly competing in terms of convenience, and as the three biggest competitors this is the core market. Looks like this territory is declining a bit, but it's the key source of market share so important short term.

Some of the adjacent, "Tier 2" brands however are growing - they are more oriented towards food and restaurant quality, so we are perhaps seeing a long term shift in the market.

This tells a lot about market structure. As correspondence analysis reduces things to two dimensions, we should check the source data to validate our conclusions and tell us a bit more.

 

Data Check: Brand Attribute Grid Patterns

As we made some changes to the Brand Attribute data to set up our map, I'll duplicate the question

Switch rows and columns, Hide 'Net' Column

Resize, Properties > Font = 9

The significant results are all highlighted. I can automatically sort the table to show the patterns more clearly.

Right Click > Sort > Categories Shown in Columns > By Pattern

This confirms our Tier 1 and Tier 2 competitors

Right Click > Sort > Categories Shown in Rows > By Pattern

We can see we are behind the leader on Drive Thru service and Convenient Opening Hours, so we have some specific things we can focus on short term

We can also see of all the tier one brands we compete with the tier 2 brands on Restaurant Appearance and Décor, so something to keep an eye on

 

Most Valuable Consumer

Next we deal with the Most Valuable Consumer, that is, who we are selling to, who should be at the center of our strategic thinking. We might want to look at market segmentation at some point, but before we do that we'll want to define the core consumer.

There are different ways to do this, we are going to focus on volume, or market contribution

We'll use a histogram look at the distribution of behaviour, using CART to see if there is a clear demographic "center" for our market, and otherwise create behavioral segments and see what we can learn about them

 

Distribution of consumption volume

We need to stat by finding or creating a variable that measures volume of consumption

I've got a variable set showing how many times consumers ordered from each brand in the last month.

It makes sense to focus on the market as defined by our Tier 1 and Tier 2 competitors, using the insight from earlier.

Select Burger Shack, Burger Chef, Nuovo Burger, Southern Fried Chicken, Arnold's, Ma's Burger

Calculate > Sum

Right click new variable > Rename > Total Occasions (key comp)

Drag across Total Occasions (key comp)

Visualization > Distributions > Categorizable Histogram

We can see here there is a long tail of consumption. Some consumers might visit for coffee, lunch, snacking more than once a day, but let's cap it at 50 occasions, keep a lid on extreme behaviour

Select Total Occasions (key comp), type "capp" in search box, Anything > Data > Variables > Modify > Recode > Recode High Values, '50', OK

 

Finding the center: Classification and Regression Trees

A good technique for find the demographic center of a category or market is CART.

Anything > Advanced Analysis > Machine Learning > CART

Outcome: 'Total Occasions (key comp)'

Now I'll select my demographics

Predictors, select A1-A3

Predictor category labels: Full labels

Here we get a nice Sankey tree visualization

We can see that the best predictor is age, with heavier users, shown in blue, being aged under 35, followed by gender, with younger males more frequent

This is interesting but at as order frequency is not that different between the groups. Let's see if we can learn more …

 

Distribution of consumption volume

We'll try another way of finding our Most Valuable Consumers

I'll use a copy of the variable, to keep our Tree intact.

Swap out row variable for Total Occasions (key comp 2)

Select Histogram, Properties > HISTOGRAM CATEGORIES > With equal proportions

Displayr has best fit 3 categories of roughly equal proportions. We can modify them by dragging the red bars.

Being monthly data, the mid point is roughly once a week, then we have higher and lower frequency segments, so I'll leave these

A variable containing these groups has been automatically created

Find Histogram Categories variable at bottom of tree, right click, rename it 'Behavioral Segmentation'

Relabel rows Light, Medium, Heavy

Drag across 'Total Occasions (key comp)' variable as the Column variable on second table; change statistic to column share.

We can see that the heaviest consumers account for 29% of occasions and more than double that in volume, so they are important consumers to know.

Hover under page, click '+', find Reports > Cross Tabs

For rows, select all demographics A1, A2, A3 at top and A4 thru A7 at the bottom, also select D2 Attitudes to category.

This time I'll add in some attitudinal data

OK, then select Behaviour Segments for the columns, OK

A very handy feature is to be able to sort these tables in order of significance, this focusses us on where the action is in the data, and gives us confidence when we start to build some conclusions.

Select all cross tabs, right click, Sort > Pages > By Significance.

We can see that Heavy users are a little male skewed, are a little more likely to be Single, with No Children, and between the age of 19-24.

However by far the strongest predictor is attitude

Select table > Sort > Categories in Rows > By Pattern

This are means scores on a 5 point agreement scale. We can switch them to percentages by changing the variable type.

Select variable D2, change to Binary - Multi, select Fix to select scores of 5 only.

We can see now that heavy usage is less about things like comfort and reward and more about convenience in the context of a busy lifestyle, with consumers replacing meals cooked or prepared at home when the value equation is right..

 

Conversion Funnel

So, we know which competitors and consumers we need to focus on. Our next framework helps us work out what to do - specifically what customer-centric goals we need to set and how we might go about actioning them.

Here we'll use classic brand health data and a measure of performance, market share, and apply tools like a summary matrix, a funnel, and conversion rates. The idea is that if we can understand gaps in an outcome variable like market share, we can set the right goals.

 

Conversion Funnel Examples

Just about all customer acquisition processes can be viewed as funnels or a series of sequential stages.

Conversion is just the ratio of the numbers at one stage over the previous stage. So, in the funnel on the left we have 90% conversion from visited last trip to being loyal. In the one on the right we have 25% conversion from being in the CRM through to having received a demo.

The key to performing an effective analysis is to compare conversion rates. You can compare them over time, against benchmarks, or against competitors.

 

Customer-Centric Goal Setting

What we mean by customer-centric goal setting is just that - expressing goals in consumer terms that ladder to a specific marketing action.

Brand A in this example has a raw salience problem, so needs to increase awareness.

Brand C evidences a product problem, and needs to improve

Brand E is well liked, but we need to increase the consumption rate, perhaps via finding new occasions

Brand F is the market leader, and just needs to defend it's position

 

Data requirement: Brand Health Measures, Summary Matrix.

Let's set this up for Burger Chef

On this page I've set up tables for Awareness, Ever Eaten, Consideration, and Recent Order. I'll adjust recent order is just the main competitive set

Select all but Burger Shack, Burger Chef, Nuovo Burger, Arnold's, Southern Fried Chicken, Ma's Burgers > Hide

We can get good insights into branding issues by viewing all the stages of the funnel at the same time.

Table > Specialty > Brand Health table

Output in pages: Select

Sort by: row 4

Show as: Bars

Dispalyr has used the brands in common across the inputs, which is what we want, the main competitors

As we have the insight on Heavy Category users in terms of the volume they contribute, it's worth checking in with them here.

Select the 4 input tables at bottom, Anything > Filter > Control > List Box filter … then selected Behavioral Segments

You can create and apply filters in Q, but these control is the one thing you can't do

Deselect Light, Medium users

I'll actually add this to a master page, so I can use it elsewhere

We can do something nicer using Funnel or Pyramid charts

 

Burger Chef Conversion Funnel

This is currently set with the data from my earlier example
Change Output in pages: brand.health.2 …

So, what does Burger Chef have to do?

It's awareness is the same as its' two key competitors of Arnold's and Southern Fried Chicken.

Trial is also high

Return to Data Requirement: Brand Health. Select Matrix and set Output to Conversion, Sort by Row 3

Return to Burger Chef Conversion Funnel

It's conversion to Consider is a bit below Arnold's and some competitors are challenging here …

Where it is in really behind Arnold's is in terms of conversion from consider to most recent order, so we need to focus on increasing our rate of purchase, which leads us to our next framework

 

Demand Creating Conditions

Most U&A studies contain a lot of detail on purchase or consumption, often referred to as the "5W's", who, what, where, when and why

The key here is to prioritize these based on how well they predict behaviour, in this case, our recent usage target

We'll use the cross-tab routine that we saw earlier, to determine predictors, and then use scatter-plots to create importance-performance-maps

 

Importance-Performance Analysis: the basic idea

Again I am sure you are familiar with the basic idea - you determine what's important to consumers and whether the brand does well. You want to have lots of things in the top right corner. Things in the bottom right corner are things you need to improve performance on.

Data Requirements: Outcome or Target Variable

Here's our target variable

We can make the analysis clearer by reducing the number of alternatives - here we've collapsed recent order to reflect our known market structure

I simply selected these brands and then "combine"
Taking insight from one framework and using it as an input to another makes the analysis that much sharper

For rows, select all demographics A1, A2, A3 at top and A4 thru A7 at the bottom, also select all occasion variables C2 through C12

This time we'll add in some of the "5W" data.

OK, then select C1 Most recent order - structured for the columns, OK

Select all cross tabs, right click, Sort > Pages > By Significance.

There is a lot here, I can give more focus by removing the tables with weaker predictors

Select all cross tabs, Delete Split Button > Delete if not significant at the 99% level.

[C2] our core Convenience Territory features several occasions of interest.
[C11] food type is important, looks like Convenience means variety
[C5] our core Convenience Territory is a mix of some consumers being alone but mostly with others.
[C4] restaurants and drive through facilities are both important for the Convenience Territory.
[A3] this is interesting, for our second tier competitors, those driven by restaurant experience, the North East area is key. As these brands are growing, we should consider having some more flagship restaurants in that area
[A1 Age] it's worth taking a look at age, as it was Heavy category consumption was skewed younger, as we saw earlier. We can see here that younger consumers are stronger when it comes to the Restaurant Experience territory.

We've learnt a bit about the second tier, "Restaurant Experience" competition, but our bigger opportunity is improving amongst our more direct "Convenience" competition.

Let's bring this to life

Importance-Performance Analysis: Burger Chef

We'll start with when consumers order.

Here's our importance measure. For performance, I created a filter to reflect Burger Chef choice on the last occasion, and set is to row %

Select Not Selected column > Hide

Being row percentages, this indicates the proportion of people who chose Burger Chef at each occasion

Now for the scatter plot
Visualization > Scatter > Labelled Scatter, resize it to take up page.

X coordinates: Table on left: Table.C2 …

Y coordinate: Table on right: Table.C2.how.ordered.by.Chose.Burger.Chef

Grid Lines > uncheck

Chart > X Axis > Axis Title: Importance (%)

Chart > Y Axis > Axis Title: Performance (Chose Burger Chef %)

Quickly create a Square shape, remove fill, size to be one quadrant, then duplicate and align until 4 quadrants are set.

Looks like we need to focus on Dinner offers

We can go back and check this for the core consumers (select all control items). Set filter to heavy.

Looks like stronger competition at lunch for the core consumer.

We can easily repeat the same operation for other data. We just copy the whole page and swap the data

Rows: C11 Food Type. Looks like we need to focus on our fries.

Repeat for C5 Alone/With others.

Looks like we do OK when consumers are alone but are missing the group occasions

And so on. We can learn a lot about what to do, and quickly

 

Our Story So Far

As mentioned earlier, while we are not quite there with a full report, by applying these 4 frameworks we've learned a lot about Burger Chef's position in the market and what they need to do to improve, and in less than 30 minutes.

Using the philosophy of a data driven approach, as we have been doing, where each step informs the next, you could use these as interim conclusions, to then investigate further. We focused on the 5W's here to see how we can drive short term occasions - you might choose to conduct a driver analysis on improving long term

 

Take On the U&A Challenge

The key message today is that a structured, data driven approach to the analysis can get you off to a flying start, and with confidence.