We continue our rotating Startupbootcamp blog series - our ten teams take turns sharing their thoughts and experiences on their Startupbootcamp adventure. This entry is by Artas Bartas, of Spockly.
The Dilemma of Measuring
Today I remembered an old joke about the Soviet army: A young army recruit is given a shovel and told by his sergeant to dig a ditch. “But how far do you want me to dig the ditch?” – asks the recruit. “Well, just start here and finish around lunch time,” – replies the sergent without blinking. To me, this exchange serves as a perfect illustration of a relationship between top management and analysts/IT departments in many big companies. Data crunchers come to C-level people and ask them “What do you want us to measure?,” to which executives usually reply “Measure anything you want as long as it demonstrates how to get more sales”. It makes for an excellent problem to explore in this blog: If you are building a social analytics tool, what metrics should it include to attract the interest of more than just the usual band of data junkies? I started by looking at dashboards created by a zillion or so different ‘social analytics companies that offer their services for anybody with a site (or two) and a few thousand dollars to burn. I then turned to one of my favorite analytics experts – Avinash Kaushik – for advice. Kaushik blogs extensively on how to stop being a reporting squirrel, forever caught in the circle of reproducing context- and action-free reports, and become an analysis ninja, capable of uncovering trends and generating data-driven insights with a few clicks of the mouse.
Do it like they do on the Discovery Channel
The simplest way to go about constructing an analytics dashboard is to give free rein to some kick-ass developer on your team. After some initial whining about the pointlessness of the whole enterprise, you will have a dashboard closely resembling the cockpit of an inter-continental airliner from the 1970s: every last metric available through public APIs will be crammed into your dashboard blinking in some garish colors at mere whiff of the wind. All nice and dandy if you are looking for Christmas lights, but how are you supposed to identify influential customers with this tool?
Well, it’s simple: just take the number of re-tweets, subtract repeating authors, divide it by the total number of followers, multiply by 100 and you have your answer, your developer might say. Riiiigggghht. If you are not convinced with the answer, you are not the only one. The mere fact that some indicator is available out there does not mean it should necessarily go into your dashboard. After all, the dashboard exists for a reason and, usually, the reason is to sell merchandise, engage readers or detect low stock, not to hog on other websites’ metrics.
You will meet a tall, dark stranger
Putting marketing people in charge of the social analytics sounds a tad more reassuring: they dress well, they use many buzz words and even when caught staring into the void they swear to be thinking of new ways to drive company sales. Marketing people also seem to know a thing or two about customer psychology. Take, for example, the book by Pamela N. Danziger "Why People Buy Things They Don't Need: Understanding and Predicting Consumer Behavior": she does a wonderful job of explaining how people differ in their shopping habits, what reasons they use to “permit” themselves frivolous purchases and how much spontaneity they afford with different types of merchandise.
Unfortunately, even a firm grip around shopping psychology does not guarantee results on the web. While physical visitors appear in the store in flesh and blood and chit chat with shop assistant, web visitors surf your shop anonymously and have little patience for those pesky questionnaires. Even when retailers manage to receive answers to their questions, it is very difficult to say with certainty whether online buyer is “impulse shopper,” “careful indulger,” or “bargain hunter”.
The problem here is that shoppers’ attitudes determining her behavior are hard to link to specific metrics, which means that traditional art of selling cannot be captured in dashboards.
Wrestling the data
I used these examples to illustrate a simple point: A good dashboard should start with specific, actionable goals (bigger sales, more videos watched) and then go on to include specific metrics linked to these goals (income level, hours spent online) thus putting data at the service of the people and not the other way round. With this in mind, we decided to get our hands dirty and dig around for some actionable metrics one could pickup from social media:
Demographics. Attributes like gender, age, place of residence are the bread and butter of the marketing industry: Advertising people have been using them for the last 60 years (at least) and look how many Martinis they could afford with that simple trick. Getting hold of this data on the web is a no brainer: It helps you put a wealth of marketing knowledge gained offline to good use on the web. The mere fact that you can separate women from men is already a huge improvement in recommending products.
Interests. We are talking about peoples' social interests here: What music do they like? Which movies do they see? What is on their bookshelf? Do they play games? Some hardcore marketers might dismiss this information as fluff, but in the world of infinite possibilities and limited attention span, knowing what kind of entertainment appeals to a person often equals selling it. Capture this data in usable format and you can extend 30 seconds visit into four minute shopping trip filled with the excitement of discovery.
Web presence. There are tons of metrics one could put into this category: The number of Twitter followers, total views on YouTube or influence scores. In our view, some of these measures are better suited for metropolitan hipsters than for-profit businesses, though it makes sense to use these metrics to measure engagement level or identify web savvy people among your customers, if your marketing strategy requires that. Just be careful with your conclusions, because, as recent study shows, big numbers does not necessarily mean great influence.
Semantics. Brands people like, topics they discuss, and sentiments they express – this data is often unstructured and, therefore, difficult to capture in a neat and actionable way. Nonetheless, taming the beast of semantic analysis is well worth the effort, especially if your customers are interested in esoteric topics and obscure brands, because knowing what these are helps you to reach out to very focused target group and, therefore, execute better campaigns all around – be it paid search, promoted tweets or holographic projections in outer space.
Obviously, not all of this data is relevant to all businesses out there, but each of these metrics’ groups addresses a clear business problem allowing users to add or remove things as they see fit and then just focus on spotting trends and testing marketing hypothesis. Now, that would be a company where even last reporting squirrel has a chance to feel like analysis ninja.
Sign up to hear more from us when we launch social analytics dashboard for web businesses.