In the presence of perfect information, there is no need for analytics. In fact, the whole aim of analytics, with all its methods, models, algorithms, and statistics, is to get as close to perfect information as possible (or, better said, as close as necessary). The analytics professional would do well to keep this aim in mind, as it has a way of focusing the analysis and getting rid of any wasteful processes.

Let’s take an example – one that illustrates well the value of aiming for perfect information. We will have a look into the field of web analytics.

Suppose we were just given access to a large set of web analytics data for a company that sells pugs (yes, the animals), and they want you to ‘optimize their website for pug sales’. The novice analyst, without any regard to what the phrase ‘optimize for pug sales’ even means, would no doubt jump right into the boundless data set and get lost in a fog of pageviews, unique visits, time on site, and convoluted navigational summaries. But this is not what is going to happen to us, because we are going to pause, consider what the concept of perfect information can tell us, and then dive into an effective session of analytics with the aim of selling lots and lots of pugs.

A Simple Question

First, we’ll consider a question. If we could know everything we needed to know in order to optimize pug sales, what would we want to know? Don’t be realistic here; remember, we’re discussing the inherently unrealistic concept of perfect information. Here’s what my pug selling informational paradise would look like:

I would know every single person in the United States of America who wants to own a pug. I would know them by name, where they lived, how much they were willing to pay, and how many pugs they would want. I would know the best time to reach all of these people, and I would know what method they would prefer to be communicated with, whether that be phone calls, direct mail, digital media, personal visits, smoke signals or carrier pigeons. I would know what motivates their love for pugs and what need having a pug would fulfill in their lives. And remember, I would know all this about each person individually. Further, I would know, by name, each person in the United States that could be persuaded into buying a pug. I would know all of the above information about them, along with what exactly it would take for them to make the decision to take home a pug, whether that be an Internet advertisement, a informational video, a seminar, a mailer piece, a discussion over a croissant lunch, or whatever else. I would know this about each and every one of them.

Phew! What a paradise! Unfortunately (or rather, we might say, fortunately), it is impossible for me to know all of these things. Getting all this information would take an infinite amount of time, and that’s to say nothing about keeping it updated. So what are we to do? We’ve taken the proverbial journey into pug selling informational heaven, only to realize we’ll never achieve it in this life.

But we don’t need to. The picture of everything we would want to know, in a perfect situation, is a useful framework to use as we start to list out all of the things that we can know.

We can’t know by name everyone in the U.S. that currently wants to own a pug. But what indications can we use to find and get to as many of them as possible? One – people searching for anything related to ‘pugs’, especially ‘buying pugs’, in search engines, will be an attribute of these people. Optimize for these terms and put up some relevant search engine advertising. Two – people looking at dog sites or pet sites, like petfinder.com, are also in the market, perhaps, for a pug. Advertise on these sites. Three – people that find your site and actually take time to browse the ‘pugs for sale’ pages or start entering in information on your shopping cart. Optimize this process for them. And on and on.

Above, we talked about persuading people to buy pugs – about knowing what it is that will convince them to make the buy decision. We can’t know this for every single person, but we can know it for certain types of people. I’d start by doing some analysis of the various past blog posts that the company has put up – which ones have been the most popular, induced the most checkouts, and brought the highest level of engagement? What were they about? You may notice that they were about finding companionship, or about giving a dog a good home, or about guarding your house from intruders (a stretch for a pug, I know, but perhaps they’d be effective at keeping away small cats). These are key findings into what is motivating people to buy pugs. (And if such a blog with diverse topics doesn’t exist, perhaps it’s time to start one. . .)

Conclusion

We can go on and on here. But I think the point is clear – start with what you would know in the perfect scenario, and work your way backwards into what you can know with the available resources and data. As you do this, you’ll notice that the individual information we knew in paradise will manifest itself as understanding certain groups of people with similar tendencies, or customer segments. These segments are the currency of effective analytics.

Please also take note that we did not actually stay within the realm of web analytics click-stream data as we were doing our analysis. This is important! Just because you have endless web analytics data doesn’t necessarily mean it is the most useful way at getting the information you need to know. Don’t let yourself get cornered – understand what you need to know, and then use the best resources available to you, whatever they may be, to know it.

Perfect information to knowable information to the data resources available for knowing it – that’s the thought process. If you do it the other way, you’ll soon find yourself spending many hours sifting through a sea of data, only to come up thirsty for insights. Don’t let that happen to you.

Done right, the first step to doing good analytics has nothing to do with math or calculations or equations – these come later. The first step is taking time to understand the problem that is to be solved, who it’s being solved for, and what the context of the situation is. Don’t let the impatient quant inside of your head take over before you’ve taken time to set the parameters of the problem.

So, for an example, lets start out with a problem people in the business world confront on a regular basis – pricing. They need to know what the price point is that will maximize their profits, and this can often be a difficult thing to pin down – a prime opportunity for analytics to step in with a helping hand.

Segments

We’ve defined our problem: people need to know how to price accurately. Now, which of those people are we going to try and help with our analytics? People selling consumer products? People selling professional services? People selling entertainment? People selling heavy machinery?

What level of understanding do they currently have of pricing and analytical methods? Are they beginners, intermediates, advanced, or PhD level geniuses? We also want to know the scope and context of their pricing strategy – are they introducing a new product or selling an old one? Are they operating on a large scale, like a Proctor and Gamble type business, or are they selling garage sale items on eBay? How fierce is the competition for the product/service they are selling? The closer we can narrow down the segment to a definable person, the better we will be able to understand and serve them.

It’s critical that we understand this target person. If we don’t, we may find ourselves pontificating about the finer points of a predictive neural network pricing model to optimize a large scale skimming strategy to someone who’s never even seen a supply and demand curve. And that would be embarrassing.

Capabilities

Keeping all of the above in mind, it’s also important to consider your capabilities and the resources available to you for designing such an analytical product. What is your expertise? What are the capabilities of your organization? What interests you? How much can you invest in the project? To quote an ancient Greek aphorism, “Know Thyself.”

Subject Matter

Then we must consider the scope and limitations demanded by the subject matter. Pricing analytics, for example, could include treatments of topics such as penetration vs. skimming pricing strategies, value pricing strategies, Conjoint analysis, the Van Westendorp Price Sensitivity Meter, commonly used pricing algorithms, data sources for pricing across the web, how to run robust pricing experiments, and so on and so on. Many of these topics could cover a few of the segments above at the same time (value pricing, for example, would help people selling products or services equally well). What about limitations? Pricing analytics is only as useful as far as the availability or ease of gathering relevant data to take action on exists; if the situation precludes this, the subject matter cannot help.

Conclusion

The key, then, is to convert the appropriate subject matter, covering the right segments, into a product that you are optimally capable of producing. Phew! I think we are discovering that setting the correct parameters for the problem to be solved is almost as much work (if not more so) as the actual data crunching part. To aid us in the understanding of this process, I’ve created a sophisticated circle chart below that provides, at best, marginal value.  But I thought was clever all the same (I drew it in about 14 seconds):

So there you have it – because the world needs another acronym, we have created the OOPSS – Our Optimal Problem Solving Sector. (Pronounced, of course, as either “oops” or “oh oh, p.s.s.”, depending on the feeling you may want to convey at any given time)

Be there, and you and everyone you know will be happy.