Monthly Archives: January 2007

Is Your PPC Incremental?

Analytical Culture, Web Analytics

Something that analytical folks talk a lot about is “the incremental”.  In customer marketing, it’s not enough to get response, what we want to know is how much of the response was above and beyond what we could be expected had the test or promotion not been done.  Typically incrementality is measured using control groups – people just like the people you are sending the campaign to who do not receive the campaign.  Then you compare the profitability of the people who received the campaign versus the profitability of those who didn’t.  The difference is the true, incremental profitability of the campaign.

Kevin over at the MineThatData blog relates that Blue Nile has decided to lower prices as opposed to chasing rising ad costs.  I don’t know enough about their business model to really comment on that action, but I do know one thing – they (and many, many other companies) are wasting Pay-Per-Click spend on non-incremental clicks.

Check out this shot.  Search for Blue Nile in Google, you get a Pay-Per-Click ad for Blue Nile, and Blue Nile has the fiirst natural listing:

Search: Blue Nile

 

 

 

 

 

 

Now someone, somewhere, back in the beginning of PPC probably had a conference speech that told everybody to buy the search for your company’s brand name, maybe for “branding” or “exposure” purposes.  But almost every web site ranks first in the natural rankings for their own brand name.  So I ask you, is there any incremental here?  What are the chances, if the PPC ad was not there, that the visitor searching “Blue Nile” would not click on the first natural listing?

What do you suppose the volume of this search is, and how much money is being wasted by people clicking on the paid ad who would have clicked on the natural link anyway?  I asked the same question back in 2003, and though the structure of the test wasn’t as pure as most hard core test & control folks would like to see, the results were so dramatic that you can’t really argue much with the methodology.  A few percentage points here or there and you still get to the same place.

In this particular test, on a high volume PPC phrase for a site that also had a top 3 natural ranking, 77% of PPC sales were non-incremental – stolen from the natural side – and would have happened anyway without the PPC link.

I will be the first to say that results are going to vary a lot by the type of site, the phrase, the target audience, the search engine, and so on.  But Golly Gee folks, has anybody else ever questioned the wisdom of “railroading” a top natural listing with a paid listing – and especially for your own brand name?  I bet a lot of analytical folks have, they just couldn’t convince the marketing folks to test it.  After all, addressing / fixing this by taking down the PPC ad means:

1.  There will probably be a few lost sales, perhaps from people who for whatever reason, always click on paid ads (newbies?)  However, profits will increase.  Some folks will not believe in this idea, it’s that whole analytical culture problem.  You have to trust the numbers, not fear them.  Unless, of course, you have been specifically told to grow sales and not profits.  Rare, but it does happen.

2.  Then there is the whole “How could you not know this was a bad thing and continue doing it for so long?” problem.   That’s more analytical culture stuff.  Failure is a learning experience in the analytical culture, not a cause for whippings and firings and demotions.  The question is not how much money was lost, the question is how much will be saved going forward.  Right?

3.  Potentially, some PPC budget will be reallocated to other, more profitable marketing activities.  This could be a problem if there is “ego spend” involved in the management of PPC.  I wouldn’t move the money out of PPC, I would just find better ways to use the existing budget.  But the potential exists for losing it for a higher and better purpose that is proven out by the numbers.  Just optimizing the system, no hard feelings, right?

PPC Marketers, if you’d like to free up budget for something else, ask your web analytics folks if they can track this test for you.  Locate a high volume search phrase that you buy PPC for ads for where your web site naturally ranks as #1 or #2 for the same phrase.  Try shutting the PPC for that phrase off.  How much click volume do you lose versus the cost of the clicks when PPC is tuned on, and what does that translate into in terms of increased profit?

And I am really not picking on Blue Nile here – search any brand name, online or off.  Go ahead.  What do you see?

Would you fail to click on the organic link if the paid link was not there?  Really?  If you think you’d miss the organic link if the PPC link wasn’t there, send me a pic of the example.

LifeCycle Marketing

Analytical Culture, DataBase Marketing

Lisa Bradner from Forrester Research called to discuss LifeCycle Marketing, which is kind of a coincidence given the Sense and Respond post below. Apparently folks are having some difficulty with implementing the concept…

The Customer LifeCycle is really just a process that you can map, just like any other business process. At each stage of the LifeCycle you have an expected result based on the behavior of other customers as a whole or in the customer segment. You measure the behavior of individual customers against the process benchmarks, and when the customer is behaving as expected, you do nothing, taking no action. If the customer behavior is “out of bounds” with the expected result, you take action. This method generally allocates marketing spend to the highest and best use. If this sounds a bit like Six Sigma for Marketing, well, you’re right, it does. You have a problem with that?

Another way to look at it is this: there is a “tolerance” band for behavior and the ability of marketing to affect behavior depends on where the customer is within that band. If the customer moves too far outside the band, it becomes impossible for marketing to really do anything at all to affect behavior. So as long as the customer remains in that band, it conserves marketing resources to take no action. As the customer begins moving towards the band, significant marketing action is triggered and needs to be taken before the customer moves too far outside the band. So you have a reallocation of marketing resources towards highest and best use, always pushing marketing spend to where it will be most effective. It’s a marketing resource allocation model of sorts.

Let’s take a simple retail example of how this works. Let’s say you look at new customer purchase behavior, and you see for new customers who make a second purchase, they usually make the 2nd within 45 days of the first purchase. So, you can look at new customers and divide them into 2 groups; those that are doing the “expected” and those that are not, based on the 45 day rule. Applying the LifeCycle concept, any new customer that makes a second purchase within 45 days of the first, marketing does nothing (inside the band). This conserves marketing resources and margin dollars that would have been lost to discounting. That money is then reallocated and spent on the customers crossing over the 45 day tripwire without a second purchase (outside the band), and since you have more to spend (courtesy of the reallocation), the programs can be more effective.

Further, let’s say that you analyze this 45-day idea looking at the marketing campaign that generated the new customer. You have only 2 campaigns and the days between 1st and second purchase is 60 days for one and 30 days for the other (average 45 days). So, the first thing you ask yourself is why is the behavior different – media, copy, offer? The second thing you ask is should we reallocate spend from the campaign with a 60 day window to the campaign with the 30 day window, which would generally increase cash flow? And the last thing you do is adjust the original 45 day trip wire to 2 distinct tripwires, one for the 30 day campaign and one for the 60 day campaign (if you keep the 60 day campaign). You are optimizing the marketing system based on the unique LifeCycle profiles of these new customers, generally lowering costs and increasing margins as you optimize.

The thing is, this is really fundamentally the same as optimizing a web site. It’s the same idea, only with different variables and more detailed data. I think that’s why many of the web analytics folks seem to “get it” and are now working on systems to automate it. I saw a shopping cart demo last week with this kind of LifeCycle profiling built right into it. You could run the profiles and execute the LifeCycle-targeted e-mails right within the same interface.

Behavior predicts behavior. If you use behavioral metrics like Latency and Recency, you can discover these LifeCycle patterns and use them to your advantage. Every marketing system, B2C or B2B, has LifeCycle processes in it. By understanding these processes you can focus resources and increase the overall profitability of all your marketing efforts.

Why are companies having troubles implementing such a relatively simple concept? I dunno, guess we will have to see what Lisa has to say in her report…but why do companies have trouble implementing just about every data-driven Marketing or Service effort? More often than not, the root cause is lack of a proper analytical culture to support the effort.

Sense And Respond Marketing

Analytics Education, DataBase Marketing, Marketing / Tech Interface

Ron Shevlin of database marketing powerhouse Epsilon thinks a new core competency requirement for marketers is the “ability to move customers through the buying cycle with a sense-and-respond capability”. This is something I often talk to people about, it’s really a subset of the “I have the data, now what do I do?” problem. Marketers are more familiar with creating campaigns based on nameless, faceless GRP’s than the behavior of real people. And that’s the problem.

I think part of the problem is in segmentation, they simply don’t understand how powerful behavioral segmentation is, how different it is than using demographics – and they lack the ability to ask for / get this information in a format that drives action-oriented thinking. The granularity of “people” as opposed to GRP’s throws them off. With Sense And Respond Marketing, or what I would call Relationship Marketing, you use the Customer LifeCycle to influence messaging which is meaningful to people based on behavior, not demographics. The behavior is the message, not the age, income, make of car, or whatever. Using behavior makes so much more sense when you see an 80 year old on a Harley.

Here’s an example. One thing that happens with interactivity is people tend to “gorge” themselves on something, get tired of it, and move on to the next experience (video games, Friendster). So you have to work very hard to hold on to them. At HSN, we used to listen very carefully to what customers said on the air and reviewed comment trends in customer service every single day. One thing we started hearing was “I’ve only got 10 fingers” which is the customer saying “you are selling too much jewelry”. At the same time, we were looking at the LifeCycle of best customers and found that most of them were fashion buyers who started buying in jewelry – regardless of how old they were or what their incomes were.

So we have customers telling us we sell too much jewelry, and we end up losing a lot of them because they get bored. But at the same time, best customers are created when someone starts buying jewelry and moves into fashion. We have a natural transition from new customer / jewelry to best customer / fashion that some customers found their way to and others did not. Knowing this behavior exists and that it’s very profitable for HSN, can we influence it? Can we get more people to make the jewelry to fashion transition with a marketing campaign of some kind?

Well, the first thing is timing. When to drop the campaign? You can’t drop it on a “date” to all customers, you have new customers coming on each day and they are going through a LifeCycle. However, the data said if the customer did not start buying fashion by the 120th day of their LifeCycle, they would probably never buy fashion. So somewhere in that 90 – 120th day after becoming a new customer, we need to hit them with a “buy fashion” message.

OK, so what is the message?  Well, we know from customer comments (and remote selling in general) that people are reluctant to buy fashion remotely because they are worried about fit. So what would be the easiest fashion item to sell a remote customer? How about something like a running suit, you know, Small-Med-Large-XLarge?

So we put together these special fashion shows geared to “no brainer fit” fashions and had them run at very specific times on the network that we could promote to the customer in advance. We dropped a very simple piece that said, “We’d really like you to try our fashions, here is $10 off, here is when to watch” kind of thing. And we dropped it somewhere in the 90 – 120 day window after the customer’s first purchase. Understand, these pieces went out every week but they went to very specific people with specific behavior who were entering “the zone” of 90 – 120 days after first purchase of jewelry.

And we literally printed money from that point on with this program. For every $1 in cost, we generated $25 in incremental (versus control) profit in the first year of the customer life, every day, day in and day out, as a higher percentage of new customers converted into long-term, highly profitable fashion buyers.

Was that a hard program to design? Not to me, seems completely logical. You have behavior, you know the customer, you have timing points, copy is simple and direct. I think Ron probably had something a little more sophisticated in mind when he wrote Sense And Respond Marketing, but the basic concept is the same (and after all, we were dealing with mainframes and snail mail at HSN in 1994, so cut me some slack!).

So why is it again that people have this “I have the data, now what do I do” problem? I suspect it’s because they may have the data, but it’s not in any kind of actionable report format that generates ideas. GRP Marketers simply don’t know how to ask for the data / can’t get the data in a format that lends itself to creating effective campaigns. And that’s a shame, because it’s pretty simple to have someone do it for you or you can do it yourself.