Web Analytics

Making Billboards Digital and Accountable

As most offline media have the limitation of not being accountable and measureable, outdoor billboards can be improved with three simple ideas:

- Make them digital: Since electricity is already connected to billboards for lighting purposed, a new digital screen can easily replace the existing one.

- Go Online: With a simple WiMax adaptor connected to those screens, you can take these billboards online instantly, which helps in managing them through a web interface from anywhere in the world.

- Make it measurable(this is the trick): With a simple speed control camera, you can actually count the number of cars passing by the billboard, and therefore start selling impressions instead of time.

Of course, you can easily argue that you don't know how many people are in each car, and whether or not they look at the billboard.

There is a general average of people per car, and even without knowing that you still know that your campaign received X cars x Y people on average. Also, you can easily assume the minimum of 1 person per car and build the numbers on that. Whether people look or not is next to impossible to measure, and so is the case with online advertising. You cannot know if people looked at an impression or not, but the important thing is that you know that 10 million impressions reach more people than 5 million impressions.

Brands and Web Analytics

This is my attempt to describe how using web analytics we can get some insights on one brands, one of the most difficult things to measure. This presentation contains information and examples on how to use the numbers to know more about the actual brand vs. the brand you aspire to be. 

Main ideas... 

Definition: the brand is basically the sum total of what people who interacted with you think about you. It is not what you claim to be. Still not sure who created this definition but it is the best one I found so far.

Quantifying the brand: analyzing where people spend most time on your website, analyzing the keywords they use to come to your site shed a lot of light on what kind of brand they think you are, and shows whether or they care about the stuff you are promoting. Even if you have more content about topic X, but people are spending far more time on topic Y, then your site is "about" topic Y. At least in the eyes of consumers, which is what really matters, which is your actual brand.



Making Web Analytics Actionable - Part 2

"Not everything that matters can be measured and not everything that can be measured matters" said Einstein. This is the main problem in the thinking behind my first post about making analytics actionable.

The assumption is that what you can control will immediately influence results on your site. There is also an implicit assumption of a linear relationship and direct correlation.

The most important things have an exponential effect after being done consistently and for a long enough period of time. For example, if you do a lot of research and reading, it will reflect on the quality of content, people will find it useful, and they will promote it for you. There is no metric that measures what you do outside your website.

So, to refine the first argument, we need to assume that all variables are held constant, and then we can correlate some actions to certain outcomes and results.

Another important thing is to know that there are things that have to be done outside the site, but will have a huge impact on the reputation and brand, thereby affecting results of the site.

Web Analytics Wednesday - Dubai

Had a great time yesterday in our first Web Analytics Wednesday in Dubai. Around 18 people attended and it was very nice cathching up with friends and meeting new ones. The presentation I shared is mainly about using analytics as the starting point, proactively, and not as a reactive mechanism. Here is the presentation:

Here it is if you prefer Google Docs. Mani also shared an interesting presentation on social media.
You can follow tweets abouts this gathering on #WAWDubai.
It seems we will be having this next week also.

Making Web Analytics Actionable

Having actionable insights is one of the most important things in any analysis you do. Ok, understood. But now what? How do we "make" our analytics actionable?

Page views increased by 15%, or page / visit went down 17%. So what?

These are the results of things that happened on your site, and caused page views, pages/visit, or whatever you are measuring to go up or down. After discovering the disaster (or the great news) you will have to dig deep and know why it happened in order to remedy the situation.

But there is another approach, which starts the other way around, preempts problems, and gives a clear action path in situations like the above.

This is inspired by a sentence that Bryan Eisenberg said in a webinar,"It doesn't make any sense to measure anything if you don't know what you are going to do with it."


This approach starts with the available actions you can influence in your site, and then builds the measurement strategy based on that:

1. Start by asking,"what actions are available to me on the site?"

Possible answers: I can change the content, I can change the layout of the page elements, I can change PPC bids, etc...

2. For each action, list all the possible things it can affect so that you have a ready action list during analysis.

Possible examples: Adding/removing keywords from my PPC campaign affects my conversion rate, getting high quality links affects my position on search results for keywords X,Y, & Z.

3. Build you KPIs based on the things that are affected by what you have control over.

Possible examples: Conversion rate of campaigns if you have control over, pages/visit for traffic sources you can control (like PPC), user experience of a process that you can influence the business rules.

With this approach you almost automatically know what you need to do when KPIs tell you something, because you know how they are affected, and because you already chosen the ones that you can influence.

The answer to "how to make analytics actionable?"...

You don't. You see what you are already empowered to do and analyze accordingly. Important things that you can't control should definitely be reported and you should seek to influence them, but in the current situation you should start with what you have.

Frequency vs. Importance

If you look at the statistics of your mobile phone calls, you will probably realize that the frequency with which you call people is not evenly distributed.

You probably call some people every day, some people every week, and so on.

But how often do you call your mechanic? Or an emergency doctor, or the police? Probably once in the lifetime of your phone? But that one call, that brought in the doctor at the right time to save a life, or that mechanic that came helped you when you were in the middle of nowhere, was as important as all the calls you made to an "important" person combined.

Just because you don't call them frequently does not mean that they are not important.

Likewise, when analyzing visits to pages on a website, focusing on high traffic pages might cause you to loose sight of other low traffic, but potentially very important pages in terms of content. The same applies to your Facebook friends, Twitter followers, and simply, everyone you know.

AdWords an an Analytics Tool

The new AdWords interface is truly a great improvement on the product. Not just in terms of visual appeal and presentation, but especially with the set of tools and enhancements it comes with.

I'm going to focus here on one of these improvements, which is using AdWords to analyze the performance of your keywords, ad groups, and campaigns right from within your account.

Starting AdWords Campaigns: The Alternative Path of Landing Pages

Many online advertisers have a problem getting their ads to show on AdWords, or getting a reasonable CPC, instead of $5, or sometimes $10.
While there are several possible reasons for this, I'm going to emphasize the importance of selecting landing pages in the initial process (as opposed to having it as the last decision in the campaign).
The landing page is after all the first encounter your user has with your site. This is where the selling happens. This is where the branding happens. And this page is visited by a tiny fraction of users who saw your ad, users that you pay for each click they do.

Forget AdWords.

Go to your analytics tool, and check out the landing pages, or most popular content report. Sort pages according to bounce rate (or conversion rate, or $ value), and see what your top performing pages are.

This report shows you the pages that you know based on statistics that people will be less likely to run away from your site once they see them.

If your site is content based, and you have many content categories that you want to promote, you might promote those that are already performing better than others (instead of promoting the ones that you think should be promoted).

Sometimes, it is actually the very beginning of a campaign. If you discover that a certain page is giving huge results, why not create a campaign especially for it?

The Good Customer, and the Not-So-Good One

How do I classify my customers / clients? An interesting question that always arises in discussions of segmenting our client base. Whether it is types of customers, or good and bad ones, there is always a problem of drawing that "fine line". We usually know that we have clients who pay more than others, and we know that we should therefore treat them differently, but "how do we draw the line?" could become a tricky question. I prefer getting the answer from the clients collectively. I would like an approach that fits to my special case, and can be used over and over without having rigid lines differentiating between "good" and "bad" clients.
This is a simple technique where you just plot your clients on a graph, draw a line, and that's it!
First create a list of all your clients, and next to each one the average purchases they make, and rank them from the highest to the lowest.
Then, plot the results on a graph, and you will end up with a "long tail". Draw a line to separate the head and the tail, and you get a fairly good segmentation between good and bad customers.
long tail.bmp
The important thing about this technique is that it is flexible and scalable to any business, and to any number of clients. It can also be used for any time range. Also, you get rid of rigid classifications (a good customer is someone who buys more than $1,000/month). This way of classifying could become ridiculous in six months, since your business can grow and your clients' purchases also.
This classifications doesn't look at amounts, it looks at the relative positioning of your clients according to their performance with your business.