Mapping Your Data on the Cheap

One of the challenges the analytics community needs to work on is finding better ways to present our findings so that they are easy for business managers to digest. We tend to show up with tables of figures and complex spreadsheets, but that’s not always the best way to get the point across. One of our clients is in a business that is heavily geographically. So not only do they want to know how effective their campaigns have been from an ROI standpoint, they want to know this by state, or even county.

You can imagine what a spreadsheet listing all of the counties in the country and their associated metrics would look like. What a beast. Much better to show this overlaid on a map, don’t you think? Well, I searched for months to find a good, easy and cheap solution for rendering data on a map, and surprisingly came up with few options. There are sophisticated applications/services out there that can do this and more, but these were generally built for companies in logistics related industries and don’t offer the functionally needed by a marketer. What’s more, they are not cheap.

One option I have been using for 10 years now is Microsoft’s MapPoint. Now don’t misunderstand me, this is far from an ideal solution. It is clunky, limited in its options and definitely a canned product Microsoft just threw out there with little support. That said, it costs less than $300 and is very simple to use. I generally describe it as a map that allows you to import a spreadsheet. And as long as your spreadsheet has at least one column with a standard geographic metric in it, you can plot it. It is great for plotting store locations, DMAs where advertising was purchased, zip codes that convert most…or all three on the same map. Now that’s a visual!

Map of retail locations and surrounding, segmented zip codes within 20 miles

There is one quirk that I researched for weeks. It drove me crazy. There is no intuitive way to render the value of what you are plotting as the label. In other words, if a particular county produced 500 sales in a given month, you want the county name and the number “500” overlaid on the map. Let me save you hours of aggravation and tell you the workaround. Territories. When importing your spreadsheet, use the Importing Territories Wizard and assign the heading Territory to the column you want to serve as your label – “Sales” in this example.

Otherwise, the tool is a great bang for your buck. I use it all the time, and the maps it creates never cease to impress.

Site Overlay that Works

One of the sexiest reports in any web analytics application is the Site Overlay (Google Analytics) or ClickMap (Omniture). It’s nice because you don’t have to be a data jockey to get the point. This report will show you where visitors are clicking on a particular page of your web site, and what results from those clicks. What great insight if you are trying to see if your links are set up well for user experience, or if specific content is compelling to visitors.

Unfortunately, I have never seen an implementation where this report has worked properly. Even if you are careful enough to ensure that no two links on a page share the same exact path (which I almost always find), you will still likely have inaccurate reports. This is due to the technology of these web analytics applications.

But there is another option. Third party services are able to provide this insight, and can do it in more depth. Two top vendors are CrazyEgg and ClickTale. Like your web analytics applications, they simply require you to drop a little javascript on your pages, and they will provide site overlay, heatmaps and more. Great eye candy for executives and data geeks alike.

ClickTale mouse move heatmap

CrazyEgg confetti heatmap

Although these third party applications charge for their services, they are priced quite flexibly so that you can limit your spend if you are careful. I would suggest choosing a particular portion of your site and starting there. Besides, with the insight they can provide, these services can pay for themselves in no time flat.

Tips When Testing Your Next Idea

How familiar is this scene to you:  There are 10 people in a conference room.  Line-of-business owners, representatives from the creative team, a senior exec or two, and maybe even an analytics person.  The goal of the meeting – decide what to test in the upcoming campaigns.  On the table are new circulation lists or media properties, different creative executions, alternative calls-to-action, and perhaps more.  So what to do and how to agree on the next step?

In this situation, I always focus on two key points:  1) if testing discipline is not practiced, the test could be wasted, and 2) prioritize what you want to learn.  On the first point, I am speaking of ensuring you achieve a valid test result.  If you cut up the test into too many different cells, you may end up with Parameter A yielding 31 conversions, Parameter B showing 28 and Parameter C netting 35.  So after weeks (or months) of planning, execution and budget outlay, we can’t say whether the differences in performance are true or simply due to normal statistical variability.

Easy for me to say, you may be thinking.  Actually, it’s easy for you to do.  All that is necessary is doing some simple math to estimate what your results will be.  Look at previous campaigns to get a ballpark for response rates, click-through-rates, conversion rates and the like.  Then, do your simple math:

Circulation (or Impressions)  X  Response Rate  =  Total Responses

 

Next,

Total Responses  X  Conversion Rate  =  Conversions

 

Look at this figure for your different test cells.  Be sure that you can get to a count of at least 150.  The best way is to actually back into a statistically significant result, but in lieu of that at least ensure that you have this many conversions.  That will often lead you to a readable sample size and result.

Ok, so we have whittled down the number of test cells through the exercise above.  Now, how do we choose which parameters to test?  That is the tough part.  Still, as you might expect, it’s about prioritization.  I suggest you start with the boldest thing you are testing.  A good example of this is a new set of media or a new list.  Once you have validated the first priority, then move to the next.  Yes, this will take some time and in today’s world, everyone wants answers now.  But the risk is trying to do too much in each test and not learning anything.  I’ve seen this play out more times that I can count.

DRTV – It Really Comes Down to the Media

I have run across the same challenge in pretty much every organization I have encountered that advertises on television.   But here is an insight that I have seen rocket companies – most notably Rosetta Stone, who aggressively executed DRTV – to the next level.  The natural belief is that your best chance of success is to air your spot at a place and time when your target audience is sure to be tuned in.  It makes sense, so it’s hard to convince people that there is another way to look at it.  But there is.

The fact is, there are many types of media out there – the most expensive of which is guaranteed on a particular network and time.   Let’s just do the math, comparing two different media buys:

	MEDIA BUY 1	MEDIA BUY 2
NETWORK	CNN	AMC (Remnant)
TIME	Wed 8pm	Sat 11am
DURATION	60 Sec Spot	60 Sec Spot
COST	$7,000	$150
CALLS	100	15
CONVERSION	10%	7%
AVERAGE ORDER	$250	$175
REVENUE	$2,500	$184
ROI	0.36	1.23



As is shown here, a much stronger ROI can be achieved on significantly weaker performance, as long as the media cost is low.  This is not a fictitious example, but a real life case of two specific media buys.  Now, your business might be very different.  You may not have a one step sale, but are instead capturing leads that you hope to convert at a later date.  Or you may have a much higher average order size.  But in the end, the concept holds.  Even though it may be counter-intuitive at first, when you can buy media at a fraction of the cost, it may not matter whether it hits your target demo all the time.  The math of remnant media is very compelling.

The Powerful Relationship Between TV and Search

Search (or Pay-Per-Click) advertising tends to be one of the most efficient advertising channels for companies.  If you have a good product, people are finding you online.  And when they do so by Googling your brand name, they often convert well.  This is because they are already a few steps down the path.

TV advertising is usually thought as the antithesis of this.  People are often unaware of your product or brand when they see the commercial, and its cost is substantially higher than that of a Search ad.  So companies often find it hard to justify spending on TV.  But if you have felt this, let me point out one important – and often overlooked – effect of TV…

TV advertising raises awareness of your product and brand. Just to be clear, I do not mean this in a fluffy brand awareness sense, but in a direct response sense.  Let me explain.  Say you are maxing out on your “branded” search campaign.  This is likely your best performing ad group.  As in all advertising mediums, you can’t buy any more inventory than is available.  However, in Search, it is possible to create inventory! How?  You guessed it – TV.  When a spot airs, you drive people to the web to learn more about your product, and they will do it using your brand name.  So although TV can be expensive, one of it’s benefits is that it creates media in what is likely your cheapest and most efficient channel – Search.

For Startups: Do Your Math and Start Out Small

I will be participating on a panel this Friday at Digital Capital Week in DC on a topic entitled Search for Startups. Given that, I thought I would do a post directed at those of you at startups or considering starting something.

There are many ways to drive potential customers to your web site or call you, but the truth is you will never know which will work – or how well – until you try it. Sure, you can look at a competitor’s efforts or a campaign you did in another medium to get an indication, but that will still simply be an educated guess. So my advice to you is:

1. Test many things
2. Start small
3. Prove effectiveness first
4. And only then…scale

This may sound obvious, but too many first timers get blinded by the concept of driving hoards of traffic – seemingly easily – by setting up a simple AdWords account.  And once they have blown their budget, they start to realize they should have been more disciplined.  Don’t make that mistake.  I have run hundreds of campaigns, and I am still sometimes surprised when I see the results. That’s the point. You really can’t know for sure, so test it and tread cautiously.

But just as important is doing your homework first. Before you even contemplate spending that first $200 on pay-per-click or a newspaper ad, lay out your funnel to see if there is even a snowballs chance that you will see a positive return.  How? Start with the size of your audience or impressions, then walk through clicks or calls, then conversion and finally average purchase amount. Begin with industry standards for conversion, and be conservative with your assumptions around clicks and purchase amount. (There are tools for this too, like Google’s Traffic Estimator).

This exercise will produce an estimate of the cost vs. the revenue of a campaign – your ROI. Of course this will be wrong, but hopefully it is in the right ballpark – and it will dissuade you from doing something foolish. Note: for those of you in recurring revenue or repeat purchase businesses, there is more room to work with because your initial sale does not need to be ROI positive, but you should definitely model in repeat purchases. Just be conservative in your assumption.

There’s plenty of time to dial things up once you find a winning formula. For now, focus on getting a little in-market data while not losing your shirt on your maiden voyage. Good luck!

Stepping Into Broadcast Advertising

Your business is growing steadily. Things are going well. You started out on Google bidding 50 cents here and $1.30 there and built a stable channel in PPC. You then branched out to the affiliate world and navigated that questionable lot to find some good partners and grow your business even more. Then you took a leap and placed your first print ad, and held your breath for 2 months until you learned that it worked. And today, although you have a solid base of advertising in these vehicles, you still want to grow. But you pause before you take the next step because that is a whole different beast – broadcast advertising.

Broadcast advertising doesn’t have to be scary, but there are some important considerations before jumping into it. I am going to weigh the pros and cons of Direct Response Television (DRTV) and Direct Response Radio (DRR) here, but these principles apply even in the brand advertising world.

So your first question may be, how do I ease into this? This is the first pro for radio. Production for a radio spot is minimal, as it simply involves people speaking and a few sound effects. Even with numerous takes, it requires minimal time in the studio. Just make sure you get someone good to write your creative, because it will definitely impact performance. Next, you need to place some media and test the channel. This is again a positive for radio versus TV. For the price of one to two full pages in a monthly you can test radio and get a decent read.

Based on this, you may be thinking that radio is the obvious way to go. However, that may not be true. Yes, it will cost you significantly to produce and test a TV spot. But TV has the potential to change your business – as it has for Rosetta Stone and many others. Radio has done that for a handful of companies, but more times than not, it is an average performer in a company’s advertising portfolio. There are several reasons why, but the two biggest are 1) lack of focus and 2) difficulty scaling. What is a lack of focus? Think about when you listen to the radio versus watch TV. Usually you are doing something else while listening to the radio (driving, cleaning, working). This makes for a less attentive audience, and one that may not be able to write down a phone number or web address while driving. But the second reason is even more significant. On TV, there are maybe a few hundred networks. If you whittle those down to the big networks that have significant media available, we are talking dozens. But with radio, there are literally thousands of stations. Centralizing, managing and aggregating data for all those stations is a bear. There are a few agencies out there who do this, but not many good ones.

TV, in spite of its high hurdle of initial investment, can be a game changer. Now, it’s only for companies that have the cash, are patient enough to see it through (no spot hits it out of the park on the first go), and are able to handle real growth. As well, TV really only works for certain types of products. But if all these things come together, it can be an amazing medium. Just make sure you pick the right agency, because another downside of TV is, there are dozens of players that will all tell you whatever they think you want to hear. They get compensated based on the media you buy, so their interests may not be fully aligned with yours. You should be very critical when interviewing agencies.

Unraveling Multi-Channel Marketing

If your company advertises, chances are you are a “Multi-Channel Marketer”.  Buzzwords aside, all that really means is that you try and reach customers in several different ways (for example, Google AdWords, magazines, direct mail and email).  So why all the fuss about Multi-Channel Marketing?  The fuss is not about the concept, but how to measure and optimize it.  That means answering questions your management team has likely asked, such as:

With us advertising in so many different ways, isn’t there wasteful spending going on?

or

If we hit the same customer several different ways in the span of a month, will they convert better?

This is the billion dollar question in advertising today.  The truth is, optimizing your mix is very challenging and there is no one single way to do it.  I have heard numerous agency presentations on how to do this, and have rolled up my sleeves and tackled it myself.  The approaches are varied, but there are two that are the most common.

The first is to conduct an in-market test, identifying like populations and isolating different media into a series of test cells.  I conducted my first of these analyses at Dell in the early days of marketing analytics in 2002.  It was an enlightening exercise, and yielded some powerful recommendations.  Of course, not every company has the resources (or patience) to execute a test like this.

That brings me to the other common approach, which is a mathematical exercise.  Data from all your efforts are entered into a database and correlations, etc. are run to determine what happens as your different media fluctuates over time.  This approach may yield interesting results, but it is certainly not a silver bullet.

After nearly a decade of trying to answer this question, where I net out is that there are several approaches to improving multi-channel effectiveness, but each situation is unique and thus each approach must be as well.  I don’t believe an outsider can come in, plug your data into a computer and produce the answer.  Instead, the approach that has worked for me is to take the time to understand what really makes the business tick.  Each business has its quirks, and I have learned that this path will lead you to a better solution.

Coke and the Law of Large Numbers

I just saw a TV spot by Coke, focusing on their green ways.  I fully support the environment, but I got a chuckle out of this one.  The spot said something like If you have had a Coke in the past 30 years, you have been part of the largest recycling effort in history.  On the surface, this may seem impressive.  But my mind immediately jumped to, what does this mean if we drill in a bit? Well, let’s think about this.  Coke is one of the most consumed drinks on the planet and it comes in plastic and aluminum containers.  So even a minor effort to recycle within this huge population of recyclables would have a pretty big impact.

Don’t misunderstand me.  Coke may well have been putting a lot of effort into recycling, and if so they should be applauded.   But given how much recyclable material they distribute globally, they could recycle much less on a percentage basis than most companies and still dwarf them.

Either a job well done, or a cleverly worded ad campaign.

Do You Really Need More Data?

I was reading about Seth Godin coming to town, was reviewing his collateral for this “roadtrip”, and a headline struck me.  It said “You Don’t Need More Data”.  You may be surprised to hear me say this, but I generally agree.  Let’s be honest, you probably have plenty of data.  Whether it be web data from Omniture or Google Analytics, customer data from your CRM, SKU data from your sales database, or call center data from your switch, you have access to valuable information on what your customers preferences and choices are.  There are undoubtedly some gold nuggets in all that data.  So before you go out and spend a lot of money on a new data feed or application, perhaps you should revisit the data you already have.

Truth is, most organizations I have been associated with have more data than they know what to do with, but an understaffed or underpowered analytics team to make sense of it.  This is really a missed opportunity.  And many companies that do have analysts on staff focus their time on generating dashboards and easily digested bites of data.  But the real juicy stuff is deeper than at the dashboard level.  What you ideally want is enough bandwidth for the analysts to do what I call swim in the data for a while.  This is how you uncover trends, correlations and other important patterns.

Just today, I was just in a meeting with a software vendor who recounted an experience.  The story was about Budweiser analyzing their sales at 7-11 convenience stores.  Apparently, a predictive modeling algorithm had indicated that they would see a 75% increase in sales if they placed their beer next to diapers on the shelf.  They did it and saw an 86% improvement!  This is not the kind of insight a dashboard will give you.  So rather than spending money and time generating more data, perhaps focus your limited resources on making better use of what you already have – and give your analysts the latitude to play a little.  I am betting you will see some true gains from that exercise.