This is a long post brought about through recent meeting with PeekYou, the massive database of public information on 250 million individuals, mostly in the United States, that PeekYou has managed to assemble and map back to social profiles, blogs, etc.     I’m also just about the very first person to be shown the case study I’m about to share aspects of here.

The case study I’m sharing  is on  Starbucks and was unofficially published this month, yesterday, in fact.

First things First

PeekYou is like a massive DNS server (Domain Name Service) but for people, identities, that is.   PeekData is a enterprise service that takes all the information in just about every important social media platform and organizes it into their own database that is then made available to third party applications (like Social Media Monitoring Platforms – think of the possibilities by considering platforms like Sysomos, Radian6, Brandtology, Alterian, ScoutLabs and many more as suitable third part apps for this kind of information) some of which are probably underway now (my guess).

PeekYou also uses an additional service in this case study called SpatialKey that takes data they have assembled and quickly maps it (with a few minutes).    I’m not sure if PeekYou created SpatialKey or if they license it, but it’s part of PeekYou’s mapping features and functionality.  You can get mappings based on demographics, lifestyles, school affiliations, occupation, Social Profiles, Market Schema and more.

The Starbucks PeekYou Study

There’s a lot of information to share and I can’ t put it all in this post – I will share what attracts my attention now, though.  I’m not sure where all the Starbucks data came from (was it all from Twitter or is it a mixture of a number of sources)?

PeekYou aggregates a lot of data, not all of it that can be resolved to an identity or a set of interests, but a sizable amount (30%-40%) can be positively identified while another 30% can provide information as well, just not as much; the rest can’t be identified.

The numbers above aren’t that surprising – of the entire data related to Starbucks PeekYou managed to identify positively 14,356 individuals via their social profiles, almost half of them are struggling on between 30K-60K a year income, split between 4 age groups, etc.

Note: PeekYou mentions demographics alot  ant they believe demographics overlays to existing data are neccessary to more fully understand it. The Demographics overlay is the main premise for the case study.

My own feeling is while the demographics data is really interesting, I’m not sure demographic data is as much  a predictor of anything as behavioral data is.  But it’s also true that people who come from the same area will often share some predispositions .  In small samples of data, those demographic predispositions are next to useless but for a bigger site they can become very useful.

What can we learn from the PeekData pulled from Starbucks profiles?

Depending on what you want to do the insights can be very helpful for marketing intelligence  – defining what to market to whom and when in this case.

PeekYou managed to find 583 people who are Starbucks fans who live in Manhattan, and where they live by neighborhood.   I can see all kinds of possibilities where sites that have large twitter followings could use the data – even celebrities could use it to determine where their fans are and therefore schedule appearances there.

The same thing can be done with sentiment (mapping); I like the idea but with all the known problems scoring sentiment analysis, as it is, I’m not too trusting of these results.

On the other hand, if this worked and if the sentiment was accurate enough, the information could extremely helpful (ie: for a political candidate, etc).

The Temporal Heat Grid is my favorite chart in the Starbucks case study and reminds me of some of my own visualizations, but this one is better than mine.

So, people who are Starbucks fans tend to like Tuesday and Wednesday mornings to come in, drink coffee, etc.

Finally, PeekYou can take it’s “PeekScore” of Social Media activity per individual and chart it against the Starbucks group.

I’m going to say something about PeekScore – I think it’s more interesting when comparing different groups and finding out how active members are in one group compared to another.  Otherwise I think PeekScore is a useful, if flawed used to identify influencers but the problem with this score is there are many people in the world who are influential but not active on Social Networks.

At any rate, if PeekYou were to turn their engine on a site like Comscore, used the same categories, then compared interest and influence across each segment, I’d like it – hint hint.

To take away a summary from all of this – Social Media Monitoring platforms are bound to get better soon - a few that I know of are playing with PeekYou’s data right now and it’ s likely we’ll see several integrations next year.  When that happens the influencer aspect of platforms will improve drastically (think Radian6 and Sysomos, to name a few) and while you will not see everyone’s profile who talks about a subject, you’ll be able to identify enough of them.

A even more powerful use case of PeekYou/PeekData Reverse Keyword Lookup API

But the more I think about it, the more I think most powerful use case is back link analysis of a website or web page.   As it stands now, anyone that has a list of urls and wants to know the identity of those urls (the people behind the website/page) will benefit for PeekYou’s Reverse Keyword Lookup API – but the issue with Social Media Monitoring Platforms selection of urls that is culled by Radian6, Sysomos, BrandWatch, Crimson Hexagon are all keyword based.

Never mind PeekYou – the urls you get are going to be more or less useful based on how good the keywords you used to construct the query – or how well you have targeted the tools to extract the urls/pages your interested in seeing.

On the other hand,  a website or page is a fairly static affair – using PeekYou to resolve the identities of people who have linked to a page is actually more useful. In theory you would think that Google has all that information already via it’s Google Accounts database – but the truth is, they don’t have all the data you think they might have.  For as much data as Google has, I think it’s in the dark about the identity of many of the urls in their search index.

I’m suggesting that PeekYou attack this problem by using Backlink analysis to query popular pages and tell us who the audience is for that page. The same thing could be done for newspaper articles and they can be done for membership sites and any site or page that essentially has a lot of links.

PeekYou  should continue to pursue Social Media Monitoring platforms but keep an open mind about other uses (such as those I suggested) like page backlinks.