Nano-Analytics: The minimum threshold

I challenge you to refute the following: there may be no truly useful involvement of WA for micro-businesses with no pre-existing data or even market, and that to claim so would be relying on very noisy data (i.e. that of a few testers).

How do you permeate "data-driven insight culture" into the nano-scale?

-Maciek Adwent in the comments section of the last post

Analytics is defined as the application of statistical methods to data to derive business insights.

Without data, there is no analytics.

QED

I'll argue that the definition of a business is the taking inputs, adding value to them, and the production of outputs with the intent to sell them. A profitable business takes inputs, adds value to them, and sell the outputs for a profit.

Where there are inputs, and where there are customers, there is data. At bear minimum, in Canada, a business must maintain a spreadsheet that lists inputs (usually in the form of expenses) and outputs (usually in the form of accounts receivable invoices). That raw data can be analyzed. Even if a business has only free inputs (the internet) and only one customer, there is still a single invoice. That's not much to go on. But it still makes analytics possible.

If we follow the Novo line, a line that I repeat frequently when talking about restaurant and hair salon analytics, small businesses can certainly benefit from analytics. The problem there isn't even threshold, it's demand generation (marketing the marketing analytics).

At the nano-level, the threshold for analytics is data. So, there is a minimum threshold.

The 2-Man Shop Counter-arguement

"you need to rework this morning's blog to apply to a 2-man startup shop. They've zero time/budget for navigating data oceans" - A friend by way of twitter

I'm not one of those people that's going to say every decision requires a data input. It doesn't.

We make hundreds of decisions a day when running a business - there literally isn't enough time, even with very efficient data access processes - for any organization to consider every single input. Just as your brain is programmed with a self-preserving ignorance - it doesn't process every single input, actively, a two man shop would operate in much the same way. Ignorance can be optimal sometimes. The laws of Recency and Anchor and Adjust are at play.

I'll say though that that two man startup shop should be consulting data when making huge decisions.

For instance, when making decisions on entering a market, or pursuing a client - analytical inputs and methods are important. When you're deciding the shortest path between your current location and a port, it helps to have instrumentation and a map. You wouldn't navigate from Panama to New Zealand on 'gut' alone. You'd certainly use your gut to see and avoid that massive wave off of port.

The world is not filled with statisticians or web analysts, and it's irrational to expect everybody can or will become somewhat educated about analytics. You don't want everybody to be analytical. But, you'd certainly want that two-man shop to have the knowledge that they can reach out to an analytics specialist to confirm or reject their gut.

I hope that clarifies.

A succinct case for measurability and concrete goals

Much of my early career is rooted in a simple, but powerful observation: organizations that set concrete targets perform better than those who set relative or abstract ones, regardless if those concrete targets are actually met. (Amazingly, as far as I could tell at the time, nobody else had made that observation in public policy.)

A clearly defined Key Performance Indicator (KPI) with a clear, concrete goal, can focus organizational energy towards that goal, and help an organization optimize their efforts.

I champion the notion that Key Performance Indicators (KPI) are the dependent variables that matter to the businesses direct goals. This means that all other 'metrics', all 4 million of them out there, are independent variables that may, or may not, have a causal effect on those dependent variables.

This is the key. This is the key to navigating in the ocean of data. You can tell, with a properly set up web analytics tool, an analytics system, and a statistical package - which variables matter. Which variables are predictors of future behavior. Which variables bring greater clarity to how the business is performing.

Bringing this together is, at present, not easy. But it's worth it.

Without any sort of analytics tools and strategy, an organization will have a harder time focusing on goals and a much harder time optimizing.

It's like sailing in and out of port through fog without instrumentation and a GPS. It's not impossible to navigate without instrumentation. I just wouldn't do it.

Would you?

Raw Data is a Commodity

Raw Data is a commodity.

That's the overwhelming conclusion I'm running into - that's the direction of my thinking over the past 4 weeks.

I had an excellent talk today with Jennifer Day. It's a 'catalytic' talk. She called me inquiring about a tweet on the pre-click analytics side, and she very patiently listened, in great detail, about the procedures involved and the value of that type of analytics. Somehow I spun off into a rant about data. (Hard to imagine).

I said, in effect:

"See, the problem with the web analytics vendors today is that they're in a false trap. They are only as successful as the people who use their tools. And, so many of the people who use their tools are not statisticians. There's a huge amount of processing power that goes into real analytics. We're moving big amounts of data around. If you're offering a distributed computing service, like web analytics, you want to minimize those processing cycles. But in the meantime, you force people like me to extract, transform, and load the data into a desktop client and run the heavy lifting analyses there. It would be comparatively easy for you guys just to do that heavy lifting within your web analytics framework, but the incentive to do so isn't obvious."

I won't quote Jennifer Day here. But, I went onto say:

"Now you have this great 'data visualization' movement. You have Google pushing it very hard with that entire flashterbation circle API thingy. Google is trying to make it easy for people to do 'analysis' by way of visualizataion, but notice how they're trying to move the processing power behind it away from their servers, and do it on the client side - on the browser. And I don't forecast browswers being able to process the kind of heavy lifting required anytime soon. They're certainly trying to get the browser to perform calculations faster and faster - but I don't know if you could do real mining that way. At least not for another 3 or 7 years."

Jennifer said a few brilliant things, which led me to say:

"Data is cheap. Data is to a business as fresh water is to a Canadian. A Canadian will go to the bathroom, deposit 15 militers of urine into 8 litres of fresh water, and flush it. Data is ridiculously cheap. Businesses are drowning in it, and there's nobody who can swim, little though detect which way the current is going."

Data is like water.

When I take the analogy too far though in hindsight - it's not pure Canadian toilet water.

It's salty. Like seawater.

Data is like seawater.

In order to get it into something consumable, it actually takes a lot of energy. It takes filtering out the fish and seaweed at minimum. But then you actually have to distill it so that what you get in the end is something consumable. There's extracting, transforming, and loading - all before analysis.

I'll be really clear on what I need to drive business value:

I really need for technologists to make the process of pointing a hose at the ocean easy. I need a desalination plant that works well. I need piping.

I need that big old Cloud Computer to actually want to crunch through my data. I need to have vendors *want* to provide me with analytical utilities.

So my note to vendors is pretty simple.:

Data is a Commodity. Stop thinking that you're selling me water, because you're not. Help me move it and use it.

Anchor and Adjust

The Internet penetration for Canada is 84%%

The landing page bounce rate for site A is 44%.

Is the bounce rate for site A unacceptably high?

Your opinion is heavily swayed by the previous fact, whether you're conscious of it or not.

The tendency of humans to anchor and adjust is insidious. Even if you hear a number that has nothing to do with the question in question, it's going to color future perceptions.

Why am I so concerned about this?

Consider how it can impact business decisions:

If people get it into their heads that 40 cents is a really good CPC based on some previous campaign - that becomes the anchor. Then, suddenly, a 4 dollar keyword looks really extravagant. But based on what? The previous 40 cent anchor? What of downstream conversion? What about loyal customer conversion?

I'm saying that the phenomenon is real, and you should be aware of it.

Screw Pageviews

Screw pageviews.

Pageviews are to 2009 as "Hits" are to 1997.

Pageviews are so simple to understand. It says, "Page" and "Views". It must mean "it's the number of pages that were viewed on the website".

And it's so scalable too! I mean, somebody can ask me, "How many pageviews did this page get", and you can tell them, and people will say, 'okay'. It's a directional number. Sort of like how 'hits' were back in the day. And in a soundbyte world, easy wins. It's not all easy though. It's easy to interpret a number the way you want to.

Sometimes, people will misinterpret 'pageviews' to mean 'people' - just as people (I verbally slip from time to time) will confuse unique visitors with 'people'. (It's not the same thing.) Truth be told, I don't actually count 'people' in web analytics.

But what do pageviews really do for me? What's the value of the metric?

We're living in the age of JQUERY and the post Flashturbation era. We're in a post-page refresh paradigm. I can go to a site, spend 60 minutes on it engaging with a game or a product comparison unit, and the site owner would believe that my visit was a failure. (Time spent on site = zero, bounce rate 100%, under the pure pageview paradigm).

There are so many better metrics to be had. Just most of them havn't been deployed (properly and harmonized with the rest of the package) yet.

Did I watch a video? How long did the video run? Did I submit a comment? Did I rate a product? What else did I do while I was on the page? How many actions did I take?

Then I ask you, the community - what would you count as a successful visit?

Isn't it sufficient that I came, I engaged, and that I am more likely to return later to continue on my relationship with your site - ultimately engaging in the future? What ever happened to the customer lifecycle? Could it be that the last generation of web analytics software was built by advertisers instead of by marketers?

I'm not just satisfied with knowing "how many". I've never been. "How many" is advertiser talk.

I want to know "how many and what", and more importantly, "how many and what and anonymously 'who'". I want web analytics for the marketer. I will gladly sacrifice understandability, 'easyness', for more complexity if it means I can tell richer, deeper, more actionable insights.

For those reasons, I say, screw pageviews.

Brand Affinity

"Some Definitions

Satisfaction is usually measured as the self-report of a transient attitude based on a recent experience. Even if one is asked about one's overall satisfaction, this response is almost perfectly correlated with the satisfaction rating of the most recent experience. ASat tells you a little bit about what's in the consumer's head but not much about what they'll do in the future.

Another key metric used is loyalty. Loyalty is often defined as repeat purchases. This is a step in the right direction. Now we are examining behaviors which are more likely to predict future behaviors than attitudes. However, for big-ticket items or items purchased occasionally, measures of repeat purchases are inadequate as there are insufficient data points to predict future behavior. In addition, one also needs an indicator that would provide some warning in regard to attitudinal and perceptual shifts that might impact future behavior. Loyalty measures provide little direction for improvement until it's too late.

So, we need a metric that tells us what a person is likely to do (how he'll behave) in a situation, and what's in his head, and that also provides a means of differentiating among consumers as to likelihood to repurchase, stay with a company, purchase more from that company, and provide referrals to that company, i.e., something that reflects likelihood of future behaviors. This indicator of behavioral satisfaction (demonstrated repeat purchases and a preference for this supplier) is what we will call Brand Affinity."

-Charles Pearlman
Information Management Special Reports, July 31, 2007


Finally, somebody I agree with on this "satisfaction/loyalty" thing.

Nice to meet you, Charles.

I think we're going to get along just fine.

The Delta in Business Questions

Business questions shift over time. It's the great Delta.

There are very specific questions that can have an immense impact on the business if they were answered and executed against. These include:

"How many [repeat] customers do I have?"
"Who are my most valuable customers?"
"What do my most valuable customers have in common?"
"Who are my least valuable customers?"
"How much is it costing me to service my least valuable customers?"
"How do I attract more customers like my most valuable ones?"
"Who should I direct discounts at? When?"
"Who should I direct 'I love you' campaigns at? When?"
"Who should I just stop saying anything to? When?"
"How are customers finding my site?"
"What channels are resulting in highest conversion?"
"How are returning customers finding my site?"
"Where are they falling off my site?, Why?"
"How much are people spending on my site?"
"How can I get people to spend more, at a higher margin, on my site?"

So, yes, we can go about building a web analytics + database analytics programme that seeks to answer these baseline questions, and then exectute against those insights. The real tragedy of the commons is that frequently, we're never asked or funded to answer those questions - either because nobody thinks to ask, or because we're instantly bogged down in the impossibilities of data aggregation. (Much of that data just isn't captured by web analytics data...so the great unification just can't happen yet, and other such reasons...)

All to often, before even these baseline 15 can be answered, the other questions start to pop up:

"What are people looking when they come to the site?"
"Why arn't people posting more comments?"
"What do people think about their experience on the site?"
"Where do people go after visiting Page X?"
"Why did this campaign succeed?"
"How did my banner do?"

And these are all very great questions, I must admit. However, not every installation is set up to answer them. For instance, in many cases, the question "how did my banner do" means that we had to have set up a specific parameter especially for 'your banner', and that the web analytics software is actually set up to listen and record that parameter. If that question wasn't identified as one that needed to be answered in the first place, don't expect any web analytics software to naturally pick it up. (19 times out of 20, it won't).

Worse, you don't want to hear how long it takes to really set up campaign tracking properly. (Because it's so easy to screw up, it takes time.)

Likewise, don't expect most web analytics software, which is predicated on the Pageview paradigm (STILL), to pick up the crucial event of 'submitting a comment', and that these people will be automatically segmented by the software (they won't be). That kind of customization, while possible in a few web analytics software, requires somebody that knows what their doing to set up.

Then there's entirely different set of questions that shift the emphasis from 'hindsight' into 'foresight'.

"How much cashflow can we expect in the second quarter?"
"How many returns can we expect next month?"
"If I decide to use blue banners instead of red ones, and spend half the money I did last year, how many conversions can I expect?"
"What would be the impact to checkout funnel completion if I added step C?"

Questions that are predictive in nature are a natural extention to the great basics.

Of course, I have yet to encounter a single piece of web analytics software that can tell me, on their own, the answers to any of those questions.

When it comes to web analytics, and by extention DW/BI systems - let's be aware that business questions shift. It's the great Delta.

The only real way, in my view, to derive the greatest value from the great Delta is make sure that we're always answering the business questions that will drive the greatest business value.

Easier said than done.