Eight Leaves Media has been developing sales and marketing programs for Origami Accounting for the past year. It’s been an exciting and successful project. For the holiday season, we created Origami gift boxes for the company to deliver to its roster of first-year clients. We think it’s a nice customer appreciation touch, and they’ve been very well received. Check out the Origami Blog to see the pictures.

• Groupon Works for Business — Groupon
• What to Expect When Taking the Group-Buying Plunge — The Globe and Mail
• Groupon Saved My Business — NBC
• Groupon’s Success Disaster — Redfin Blog
• Doing the Math on a Groupon Deal — New York Times
• To Groupon or Not to Groupon: The Profitability of Deep Discounts — Harvard Business School
• Is Groupon Good for Retailers — Harvard Business School
• How Effective Are Groupon Promotions for Businesses — Rice University
• Groupon in Retrospect — Posies Bakery and Cafe
• Burned by Daily-Deal Craze, Small Businesses Get Savvy — Wall Street Journal
• Wise for Some Restaurants, Coupons Are a Drain at Others — New York Times

Another very important consideration is how to measure the success of a daily deal. Measurement is close to our hearts, and we have a post in the works to help you sort things out.

Positioning in a nutshell

Positioning was popularized by Al Ries and Jack Trout in their 1981 bestseller Positioning: The Battle for Your Mind. Their premise is that in an over-communicated world, consumers screen and reject much of the information being offered and only accept whatever matches their prior knowledge or experience.

Businesses have to adapt to this environment by oversimplifying their message and by concentrating on narrow targets, the consumer segments that are most likely to listen and respond to their marketing. By focusing, businesses can hope to find some unoccupied space in a target consumer’s set of perceptions and set up shop, if you will, at a safe distance from competitors.

The key is to be distinctive. Take Dyson for example: its line of vacuum cleaners stand out functionally (we’re more powerful and use much less energy than conventional motors!) and symbolically (the best-looking vacuum cleaner there is!). The sharper the distinction that marketing can draw from competitive offerings, the more the business stands out in the minds of its targets.

To go further, the distinction should be summed up in one word in a prospect’s mind, a word that the company owns. Volvo owns safety. FedEx owns overnight. Effective positioning is about limiting what a brand stands for and alertly avoiding the trap of trying to be all things to all people. Instead, successful companies position themselves to be one thing to some people.

Too much communication

The positioning concept highlighted the problem of too much communication, but it also contributed to it.

Consumer choice, a growing problem in Ries and Trout’s original work, has since exploded as companies have introduced an avalanche of new offerings to serve the slightly different needs of slightly different segments. Each new product or service has had to be supported and positioned by advertising that played to a large audience while appealing to an increasingly smaller one. Which leaves consumers feeling more harassed and more besieged.

But positioning demands that companies differentiate and ruthlessly zero in on targets. The approach has become orthodox for brands. But can small businesses play the same card, and, without massive advertising budgets and in the face of consumer fatigue, should they play it at all?

The principle of minimum differentiation

If differentiation is so critical to success, why are there so many similar products in the market? Consider tennis rackets, cell phones, video games, toothbrushes, or pasta as a very small list — similarities between products in these categories overwhelm differences. And why do similar businesses tend to locate in the same areas of a city? Why are there two gas stations on adjacent city blocks? Or a string of nightclubs down the length of a city street?

In economics, the tendency of businesses or products to cluster is known as Hotelling’s law or the principle of minimum differentiation. Harold Hotelling formulated this principle to explain spatial competition. He observed that consumers prefer the nearest of two options when purchasing fixed priced goods with identical features. This forces both firms to locate in the middle of the market, because a firm that chooses to locate to one side of the middle would leave the larger portion of the market to its competitor.

Spatial differentiation, or the lack of it, is a useful metaphor for quality or feature differentiation in products. Consumers choose products from one of two firms with a mix of features that are closest to their ideal mix. Some consumers prefer curly pasta; some prefer whole wheat pasta shells. (Everyone loves spaghetti.)

If consumer preferences vary uniformly across all possible feature combinations, it again pays for both firms to create products that (sort of) appeal to everybody. If instead one firm chooses a mix of features that appeals especially well to a small segment of consumers, it leaves the larger segment to its competitor.

The contrast between clustering and differentiation should be clear. Which, as an explanation of company behaviour, sounds more plausible? And which, as a small business strategy, is more advisable? Thoughts on these questions in a later post.

Useful links

• Malcolm Gladwell on spaghetti sauce
• Why Starbucks actually helps mom and pop coffeehouses
• The principle of minimum differentiation revisited (pdf)
• Too much choice leaving us bewildered and depressed

From a marketing perspective, it makes sense to put in the effort to understand the characteristics and preferences of your best customers for at least two reasons: 1) to continue to provide this group with what they’re looking for and keep them as customers, and 2) to target your marketing efforts toward prospects who resemble your best customers.

By targeting your acquisition marketing through insights into your best customers, you attract customers who are likely to respond to the strengths of your small business and remain loyal to it. Instead of moving random customers up loyalty ladders, you focus instead on getting the right customers, customers who will be loyal from the start.

But, before you can start to understand your best customers, you first need to identify them. And that’s where a simple database marketing tool called recency, frequency, monetary analysis (or RFM) comes in handy.

RFM in a nutshell

RFM uses sales data to segment a pool of customers based on their purchasing behavior. The resulting customer segments are neatly ordered from most valuable to least valuable. This makes it straightforward to identify best customers.

The idea behind RFM is quite simple: 1) Customers who have purchased from you recently are more likely to buy from you again than customers who you haven’t seen for a while. 2) Customers who buy from you more often are more likely to buy again than customers who buy infrequently. 3) Customers who spend more are more likely to buy again than customers who spend less.

The order of the attributes in RFM corresponds to the order of their importance in ranking customers. Recency is the most important factor. Why?

Because the longer it takes for a customer to return to your business, the less likely he or she is to return at all. You can fix problems with good customers not coming in as often or spending as much, mostly because they’re still coming in. But when good customers stop coming in altogether — that problem is much harder to fix.

Recency alone won’t sort out your good customers from your new ones. You need frequency for that. Frequency measures the intensity of a customer’s relationship with your business. And good customers, by definition, do business with you more often. You’re part of their habit.

How much a customer spends on average or in total is the final measure of his or her value. The M in RFM adds another level of detail to the customer picture, helping you distinguish between relatively light and heavy spenders. Its effect is often, but not always, highly correlated with frequency.

Calculating RFM scores

To calculate RFM scores, you first need the values of three attributes for each customer: 1) most recent purchase date, 2) number of transactions within the period (often a year), and 3) total or average sales attributed to the customer (total or average margin works even better).

You then have to decide the number of categories for each RFM attribute. The number is typically 3 or 5. If you decide to code each RFM attribute into 3 categories, you’ll end up with 27 different coding combinations ranging from a high of 333 to a low of 111. Generally speaking, the higher the RFM score, the more valuable the customer.

You can assign customers to categories by sorting on RFM attributes or by applying business rules. For example, customers can be assigned to frequency category 3 if they have made 10 or more purchases in the past year, category 2 if they have made 3-9 purchases, and category 1 if they have made 1 or 2 purchases. Determining meaningful rules often requires a bit of data mining, however, so it’s common for RFM users to simply sort the customer file on each attribute and assign customers to categories from top to bottom (i.e., the top third of customers on frequency are assigned to category 3 and so on).

You also have to remember to sort your customers on recency first, then sort on frequency in each recency category, and, finally, sort on monetary value in each combination of recency and frequency categories. This way, you end up with an equal number of customers for each RFM score.

Using RFM scores

Once you’ve calculated RFM scores, it’s easy to identify your best customers — they have the highest score. You can now start to analyze the characteristics and purchasing behavior of this group and try to understand what distinguishes them from typical customers. Do they tend to buy a subset of your products or services? Do they live in demographically similar neighborhoods? Are their lifestyles and/or life stages similar? Why do they perceive more value in your business than the folks who you see once or twice?

The answers to such questions help you sharpen your understanding of your target market and be more precise in communicating with actual and potential customers. It’s a systematic approach to marketing, and it’s the approach we recommend to all our clients.

I’ve tried to suggest here that RFM is an effective and easy to implement method for segmenting a customer pool and identifying best customers. But I’ve avoided talking about using RFM to predict customer response for direct marketing campaigns.

The reason for that is that there are much more effective techniques than RFM for predicting customer behavior. The tradeoff with these more advanced methods is that, from a small business perspective, they are not easy to understand or implement.

As a predictive tool, RFM works — your best customers are more likely to respond to a direct mail offer than customers you rarely see. You should use RFM to focus your mailings on the most responsive segments in your customer database. But you should value the technique for what it is: merely the first step in using analytics to improve your decision-making.

I have two objectives for this post: 1) provide a more useful list of Edmonton marketing companies; and by using very simple on-page SEO techniques, 2) try to get ranked for searches on marketing companies in Edmonton. If you found this page on Google and it helps you find a local marketing company that fits your needs, that means I’ve succeeded in both the first objective (building a useful list), and in the second (giving Google the type of content it likes). So, let’s get listing!

Marketing Companies in Edmonton

1. Calder Bateman — The big guys. Advertising, marketing, design, and public relations. They’re full service: television, radio, print, outdoor, web. Even social media. The masterminds behind the Government of Alberta rebrand. They cleaned up at the 2010 ACE Awards.
2. DDB Edmonton — Local office of global advertising agency. Often winners at the ACE Awards.
3. Donovan Creative — Creative communications. The talent behind the much-admired EIA’s “Language of Speed” campaign.
4. Eight Leaves Media — That’s us. We’re a direct marketing company. We specialize in analytics and data-driven techniques that increase marketing ROI.
5. g[squared] — One of the companies on Google’s first page. They’re running a lot of creative transit ads in the Edmonton market. Very active in the Edmonton community.
6. Incite — A relationship marketing company. Named to Alberta Venture’s Fast 50 in 2009. Also made it to Google’s first page for Edmonton marketing companies.
7. Leger Marketing — Edmonton office of North American market research firm.
8. Marketing by Design — Boutique firm. Design, print, web, signage, even 3D and animation. Lots of experience with developers.
9. McRobbie Optamedia — Strategy and creative. Winners of many creative awards.
10. mkm — Margaret Kool Marketing. Brochures, logos, print, broadcast.
11. Parcom Marketing — Partners in Communication. A seasoned marketing agency that works in print, broadcast, and web. The crowdsourced edmontonstories.ca is one of their babies.
12. RED the Agency — Formerly local, now regional advertising agency that mixes traditional and digital. Edgy.
13. Urban Jungle — Full service marketing boutique. Advertising, branding, communications, design.

Keep in mind that these are the marketing companies in Edmonton that I know of. I’ve also tried to limit the list to strictly marketing companies — which is why I haven’t included Edmonton companies that are focused on web design or Internet marketing.

1. Discover influential users on topics that interest you.

Want to know the central players in networks that are tweeting about Boston, Seattle, Ottawa or 30 other North American cities? How about political networks like the Tea Party or Gov 2.0? Check out our topic-specific ranking pages like this one for New York City:

2. Listen to what influential users are saying right now about your topic.

It’s 2:10 AM (MDT). As I write this, Calgary influentials are tweeting about #yycvote; Portland influentials are tweeting about the Farmers Market; and Gov 2.0 influentials are tweeting about the First Mass Mobile Alert System. How do I know this? I checked the topic-specific stream page like this one show below for Austin, Texas:

3. Look up detailed profiles of influential users.

TweetHeroes profiles tell you what you need to know about influential Twitter users — including you. Who are your influential followers? What are your popular tweets? Who retweets you the most? How often are you being mentioned? Where are your followers from? These are just some of the questions answered by our profiles. The profile below is for Boston tweeter Laura Fitton, aka Pistachio.

That’s just the basics. We’ve made good progress over the past few months. And we have much more in store — more topics, more features. Look up profiles, tell us what you think, and please stay tuned.

Update: Twitter Analytics is almost ready for prime-time. The features we’ve seen in the sneak peek incorporate most of what we set out to do with TweetHeroes. So we’ve decided to stop development on this project and take it offline. It was a fun detour and we learned a lot about Twitter data and social network analysis. But we’re moving on. Thanks to everyone who had a look.

There are two sides to the debate: support the plebiscite or dead issue, move on.  So where does everyone stand?

Building the #ecca network

Good question.  Because our team is conditioned to see everything in terms of networks, we decided to dig into the #yeg #ecca stream over the past couple of months to see how the players in this tempest are connected.  We identified #yeg Twitter users who’ve tweeted at least twice about #ecca and #yeg in that roughly two-month time frame.  We then built the retweet network for these users, connecting two users if either one has retweeted the other.  (Twitter retweets on issue-specific tags are excellent indicators of affinity or a shared position.)

At this point, we had a confusing and densely connected graph — the #ecca tag is very popular!  What we really wanted to know though was where everyone stood, not just on the issue, but in relation to each other.  Who was on each side?  Who was central?  Who was supporting whom?

Community mining the #ecca network

Dividing the #ecca retweet graph into distinct communities was the next step.  We turned to some core community mining algorithms from social network analysis to get this result (click the image to enlarge):

There are two large communities in the #ecca network.  The community on the left supports the plebiscite, the community on the right supports closure.  The smaller communities are also on one or the other side of the issue.  They’re just not as actively interacting with members of the larger communities through #ecca retweets.

Two houses divided

The plebiscite community has several leaders or hubs: @EnvisionEdm, @Sirthinks, @BRinYEG, and @fusedlogic. The closure community has one: @mastermaq (possibly @ChrisLaBossiere as well).  We used some text categorization techniques to see what each side had to say.

The plebiscite community’s position is: 1) the #ecca provides important services to Edmonton and surrounding regions; 2) they have a vision to transform #ecca land into a sustainable industrial community; and 3) a plebiscite is the only fair and democratic way of resolving the future of the #ecca.

The closure community replies: 1) there are other ways to provide the same services; 2) given alternative uses for the land, it’s not a good business or community decision to keep the airport open; and 3) didn’t #yeg already decide the future of the #ecca in a fair and democratic way?

What’s really going on in social media?

What’s amazing, from a purely technical perspective, is that we can discover the structure of these communities just by examining who’s retweeting whom — and without knowing anything about the content of their on-topic tweets!  Retweets, as we expected, are a way of openly (and sometimes subtly) taking sides on Twitter issues.

I’m sure #yeg tweeps know the players I’ve listed above and where they stand.  This is because they’re active participants (or at least observers) in the #yeg stream.  But, for someone on the outside wanting to understand the structure of networks forming around specific topics in social media, I think that our approach holds a lot of promise.  It’s the idea of that promise that’s driving the development of TweetHeroes.

Clustering in a nutshell

According to the Wikipedia article, clustering is “the assignment of a set of observations into subsets (called clusters) so that observations in the same cluster are similar in some sense.” The most famous clustering algorithm, and the one that we used for this analysis, is called k-means. (Andrew Moore has an excellent tutorial for those interested.) K-means is a relatively simple but powerful technique that’s very useful for exploring datasets. There are quite a few details that a practitioner has to sort out (e.g., scaling, collinearity, etc.), but the output of k-means often reveals clear and distinct patterns and helps us get our bearings, particularly with marketing data.

Preparing the Edmonton census data

The Edmonton neighbourhood data is in matrix form: the rows represent neighbourhoods, the columns successive five-year age ranges (0-4, 5-9, etc.). Each cell contains the number of people in the associated neighbourhood and in the associated age range. We replaced all 65 and over age ranges with an aggregate 65+ (senior) group. We then scaled the data row-wise so that all cells contained values between 0 (the minimum row value) and 1 (the maximum row value) for each row. With this scaling, the clustering was performed on the distribution of age groups in each neighbourhood — similar neighborhoods would have similar distributions of young and old.

Neighbourhood cluster centers

The k-means algorithm finds a predetermined number of clusters. (We selected 5 clusters based on a trade-off between model complexity and goodness-of-fit.) The center of each cluster is calculated by averaging the attribute (column) values for each neighbourhood in the cluster. By inspecting the average attribute values of these cluster centers, we can understand the types of neighbourhoods found in each cluster. (Remember that the cell values range from 0 to 1, and that they represent the relative proportion of each age group in each cluster.)

Neighbourhood clusters described

• Young Families: Higher proportion of infants (0-4) and adults in the 25-44 range (parents).
• Older Families: Higher proportion of older children (10-24) and 40+ adults (parents).
• Young Adults, Seniors: Higher proportion of young adults (20-34) and seniors (65+).
• Seniors: Higher proportion of 65+ adults.
• Balanced: Comparatively uniform proportions of most age ranges.

How are these 5 age-based neighbourhood clusters distributed across Edmonton?

Key findings

1. On the outskirts: Younger families are setting up at the very edge of the city where most of the new detached residential development is taking place.
2. Ring patterns: Older families are also in neighbourhoods at the edge of the city but not as far out. They occupy an inner ring around which the younger family neighbourhoods appear to be growing. The older family ring in turn encircles a core of senior-dominated neighbourhoods.
3. Mixed use: Central neighbourhoods attract younger adults without children. They also have a well-represented senior population. The further you go from the core, however, the less likely you are to find younger adults, and the more the senior population starts to stand out.
4. Balanced neighbourhoods: Balanced neighbourhoods (such as those in the southeast) appear to be made up of an ethnically-diverse, high visible minority population.
5. Clusters tend to cluster: The neighbourhood clusters aren’t scattered randomly throughout the city — they group together putting even more emphasis on the “flocking together” effect.

Stay tuned

Age-based clustering has managed to produce a fascinating picture of how the city has evolved and even to suggest how it will continue to evolve. You can see that, in the not too distant future, the many senior-dominated neighbourhoods in the core of the city will experience significant demographic shifts as residential property starts to change hands. Given the fact that younger families have settled in the outskirts, where they’re likely to remain for years to come, it’s an open question as to what the face of core neighbourhoods will look like when this shift happens.

This initial attempt didn’t take into account census attributes such as dwelling unit type, dwelling unit ownership, structure type, and property type status. Tackling that is next on the agenda.

Related posts

• A Marketer’s Guide to the Edmonton Census

I was recently made aware of the First-Timers report by the good folks at the Edmonton Symphony Orchestra (ESO) — where we’ve been plying our data-driven trade for the past year. Wyman’s methodology — a) customer segmentation b) loyalty driver research c) targeted and tested promotions — is chock full of the basic staples of a data-driven diet. However, some of the report’s emphasis and a few recommendations struck me as off. With all due respect to the Wyman consultants, I’d like to offer a (somewhat) dissenting opinion on First-Timers.

A symphony is a shared experience

What makes the Wyman “killer” offer killer? The report clearly states that (more than free drinks, favorite composers, or soloists) discounts are “by far the most powerful lever.” But which discount: 2-for-1 or 50%-off? That’s easy; go 2-for-1 — it brings back more first-timers and it emphasizes that a symphony is a shared experience. Consider the case of the ESO:

• A solid majority of concert-goers buy tickets in pairs.
• Couples occupy the majority of seats in any performance (excluding kids series).
• Couples are more likely to donate than individual (male or female) account-holders.
• When couples donate, they tend to give significantly more.

Couples are vitally important to the well-being of orchestras — but orchestras, in all fairness, gives something back. John Dewey wrote: “In life that is truly life, everything overlaps and merges.” For a listening audience, a concert hall performance of a symphony is alive with overlaps, of people and memories near and distant. This breaking down of boundaries is, in my mind, a kind of generosity. And when patrons attend as a couple, and return as a couple, and subscribe as a couple, they come to see each other (and those around them) in that spirit of generosity.

An online dictionary provides this lovely definition of the word: “freedom from meanness or smallness of mind or character.” Just that gives you a fleeting impression of your better self, less separate, less apart. Dr. Francis G. Winspear donated $6 million to the building of the Winspear Centre (the largest single private donation to a performing arts facility in Canadian history) “… to have a place where the people of our region can make the most beautiful music they are capable of – and share it with each other.” Dr. Stuart G. Davis donated the Davis Concert Organ to “honour the memory of his late wife Winona.” This is the society of symphonies.

You come away from a performance believing that the way you act toward each other can be free of meanness or smallness. It echoes the crushingly beautiful words of Donne: “… any mans death diminishes me, because I am involved in Mankinde.” The Wyman report, with its technical and uninvolved taking apart of the “product” of the symphony, takes no real interest in the shared nature of the experience — which may have been because the shared part is obvious (and difficult to claim as the key finding of a top-flight study). It’s unfortunate but understandable. In management consulting, overlooking the obvious is an occupational hazard.

It’s OK to ask

With respect to turning First-Timers into subscribers, the Wyman report recommends relationship building: “Don’t ask me to marry you after the first date: I don’t want to commit yet.” One step at a time is the preferred approach, sort of like the Loyalty Ladder. The problem with this view is that customer relationships don’t always evolve in a neat, linear, sniffing-around sort of sequence. Customers make leaps, and first impressions produce emotional, often inarticulate, commitments.

With the ESO audience, as an example, patrons are as likely to become subscribers after their first performance as their second or third. (In fact, it’s not until after they’ve attended their seventh single ticket performance without having ever subscribed that the probability of their subscribing starts to go down.) The benefits of turning a single ticket buyer to a subscriber are obvious: increased seasonal revenue, higher donation likelihood, and higher probability of retention. A patron can’t become a long-term loyalist without first becoming a short-term loyalist.

So, it’s not only OK, it’s important to ask First-Timers if they’d be interested in subscribing. If your situation is similar to the ESO’s, you’ll have the same success rate as you will if you wait, but you’ll have it with a much larger prospect pool. American orchestras apparently telemarket First-Timers for subscriptions within days of their attendance. That does sound pushy (to people who wouldn’t think of subscribing), but the sales side is reaching out to warm prospects who have the (hopefully wonderful) experience of the symphony fresh in their minds. (The ESO’s subscriber acquisition campaign runs mostly in the period between seasons.)

Are they likely people?

When you read about Judy’s illustrative (and terrible) orchestra experience in the First-Timers report, the overwhelming feeling you should get as an arts marketer is not “We have to do better.” It should be “We have to do better at attracting people who are not like Judy.” Wyman uses Judy’s experience (missed shows, nightmare parking, crowded bar, unknown performance pieces, and a flood of post-concert direct marketing)   to browbeat orchestras for neglecting to provide a comprehensive value proposition. But, really, why not turn the scrutiny around and take a closer look at Judy? Are her problems really fixable?

Fred Reichheld puts the idea of customer loyalty like this: 1) Some customers prefer stable, long-term relationships. 2) Some customers are more profitable. 3) Some customers are more responsive to your particular business strengths. For a discerning business, acquiring new customers is a matter of asking of any prospects, “Are they likely people?” That is, will they be loyal, will they be profitable, and will they respond?

The Wyman First-Timer report doesn’t tell orchestras a thing about which groups to target as first-timers. Instead, it focuses on the retention of whoever shows up. But, if you consider Reichheld (and we do), instead of being independent, retention is actually a function of whoever shows up. Judy may have just been a poor prospect. That she had a terrible experience is unfortunate but not altogether unexpected. Remember, some customers are more responsive to your particular business strengths. It’s incredibly difficult for a business to create tastes and habits in a customer in a first visit. The First-Timers report tell marketers “It’s all you.” But it never is.

Now the report’s recommendations (familiar repertoire, background information, enjoyable pre- and post-concert experience, easy access to the hall, and flexible exchange of tickets) are good (except for perhaps the flexible ticket exchange for a first-time buyer), but not game-changing. What’s game-changing is the 2-for-1 discount, which clearly works, but should perhaps be used more strategically.

I think orchestras could benefit from a try-us-twice concept where the 2-for-1 offer applies to the tickets for the second performance. This way patrons buy the try-us-twice packs up front. You’re essentially getting First-Timers to commit to coming back in their first transaction. I think there are all sorts of benefits here, but we’re still in the early stages of thinking this through.

Attributes in the municipal census data

The municipal census is narrower in scope than the federal census, which was last completed in 2006. While the federal census gathers data on demographic attributes such as income, marital status, and ethnicity, the municipal census is restricted to the following: gender, age, dwelling unit ownership (owned or rented), structure type (single-detached house, duplex, row house, etc.), and dwelling unit and property type status (occupied, unoccupied, etc.).

The city provides some summary reports at the city, ward, and neighbourhood levels, but these raw tables have limited value for marketers (and residents). They are, however, a great starting point for analytics (which is what we do). This is the first in a series of posts that, I hope, will provide demographic insights into the municipal census data to help fuel marketing strategies for local Edmonton businesses and non-profits. Along the way, you’ll also get an introduction to what we mean by data-driven marketing.

Correlation in a nutshell

In statistics, correlation is a measure of the relation between two variables (e.g., height and weight, demand for a product and its price, etc.). Correlation coefficients can range from -1.00 to +1.00. A coefficient of -1.00 represents a perfect negative correlation: higher values of one variable correspond exactly to lower values of the other. A coefficient of +1.00 represents a perfect positive correlation: higher and lower values of one variable correspond exactly to higher and lower values of the other. A coefficient of 0.00 represents a lack of correlation: knowing the value of one variable does not help you predict the value of the other. (See the Wikipedia article or this detailed correlation primer for more information.)

The reason I’m going over all this is because the first analysis we did on the census data explored the correlations of the number of people in various age groups in Edmonton neighbourhoods.

Age group correlation table

The neighbourhood census data provides counts of male and female residents in successive five-year age groups that start at 0-4 and end at 85+. Not surprisingly, the correlation between the total number of males and the total number of females in a neighbourhood is around 0.98. The correlations between the number of males and females in individual age groups are also high, typically greater than 0.9. From this, we decided to ignore gender and just look at the correlation structure of the total population in the various age groups in each neighbourhood. With a little help from the conditional formatting feature of Microsoft Excel, we produced the following correlation table.

Each cell in the table shows the correlation coefficient for the age groups associated with the cell’s row and column. You can see that the table is symmetric with coefficients of 1.00 along the diagonal. The shading ranges from red (0.00 or no correlation) to green (1.00 or perfect correlation). This table really brings out the patterns in the distribution of age groups across neighbourhoods in Edmonton.

Key findings

1. No negative correlations. In general, if you find a relatively higher (lower) population of one age group in a neighbourhood, you are likely to find a relatively higher (lower) population of all other age groups in that neighbourhood. A negative correlation between age groups would indicate that these groups tended to cluster in distinct neighbourhoods, and perhaps suggest that they avoid each other.
2. Birds of a feather flock together. The green band along the diagonal shows that there is a high correlation in the population of adjacent age groups. As an example, neighbourhoods that have relatively more (fewer) 5-9 year olds also have relatively more (fewer) 10-14 year olds. The same applies when comparing 55-59 year olds and 60-64 year-olds and so on. Adjacent age groups tend to cluster in Edmonton neighbourhoods.
3. Family ties. The off-diagonal green cells show the high positive correlation between the population of children (under 19) and the population of their parents (age groups from 30-34 to 60-64). Because we combined the male and female populations, we can’t make any observations about the relative ages of mothers and fathers of the various child age groups. But, in general, it appears that the population of child age groups are most correlated with parent age groups that are 30 to 40 years older.
4. Seniors and children. The reddish areas in the top right and bottom left of the table shows that there is a relatively low (or even no) correlation between the population of seniors and children in Edmonton neighbourhoods.
5. Twenty somethings. The correlation between the population of 15-19 year olds and 20-24 year olds is surprisingly low; it’s the one off-diagonal element in the table that isn’t shaded green. But we know that this is an important life stage that involves moving out on your own, so this would likely create the break we observe in the table.

Stay tuned

As I mentioned, we’re going to continue the analysis and discussion of the municipal census in a series of upcoming posts. Our first effort was relatively straightforward, and the findings should be quite intuitive. Conditionally formatted correlation tables in Excel are quick and often helpful. While you do have to be careful not to over-interpret the results (there are more powerful multivariate techniques that apply here), the approach throws useful light on new datasets. If any of you would be interested in a step-by-step of how to go about building these tables, leave a comment, and someone on the team will provide a follow-up post with the details.

Related posts

• Flocking Together: Neighbourhood Clusters in the Edmonton Census, Part 1