Back From Space.

Posted on by
Reply

I recently finished my Ph.D in Applied Mathematics here in Boulder and have returned to work with the Spot Influence team.  Maybe I should have taken more than a week off in between to travel Europe or follow Phish for a summer, but there is too much exciting stuff happening at Spot that I couldn’t wait to get back in it.  Before I let the memories of my dissertation fade and get replaced by the constant flow of new and exciting technological knowledge we use here, let me tell you a bit about it.

My thesis work focused on randomized techniques in numerical linear algebra (nla).  The goal was to develop methods that are fast and accurate, that work well on distributed systems, and that can operate on the scale needed to process tons of large noisy data. Traditionally, there has always been an amount of randomness in nla; iterative methods might use a random starting vector or roundoff error produces slight random perturbations that steer methods away from singularity, but this randomness was not fully exploited.  The basic observation of our randomized methods is to realize that if you draw random vectors from a high dimensional space they not only are linearly independent, but they also maintain a certain quantifiable distance from each other.  In other words, if you stack the random vectors into a matrix the smallest singular value is bounded from below.

Building on these properties of random matrices, we were able to construct factorization methods that only require access the input matrix a constant number of times.  In other words, if you want 50 eigenvectors or even 500, you only need to access the matrix a few times, sometimes as few as one.  This property is extremely useful when dealing with large matrices that cannot fit in RAM since it minimizes the bottleneck of data transfer from disk.  For distributed computing, where the cost of data i/o is cut down by using many machines, the randomized methods provide an additional benefit of blocked operations and bulk computation.

These methods and their suitability for massive scale distributed matrix factorization are no secret either!  The MapReduce machine learning library Mahout incorporated a variant of these randomized algorithms, the Stochastic Singular Value Decomposition, into their 0.5 release and are continuing to add functionality as they roll out 0.6 that includes a Principal Component Analysis option.  Deploying the algorithm on Amazon’s EMR I was able to compute 100′s of singular triplets of a matrix whose dimensions exceeded 35,000,000.  This is by no means a limit to the size of data that can be processed, but simply a limit to my graduate student budget that I was working on $$.

If you want to know more you can find my thesis here.  Also check out Mahout if you need to svd some impossibly large data.

 

 

Uncovering The Future of Online Engagement: Proactive vs. Reactive

Posted on by
Reply

As a business, what’s your strategy for online engagement?

Do you spend the majority of your time responding to brand mentions and comments as they come in? Or are you constantly searching for new Influencers, commenting on industry blogs, and engaging with a broader audience?

Boiling it down – does your business take a reactive or proactive approach to social media?

It seems to us at Spot Influence that the second strategy is bound to yield better results than the first. To put it in a strategic metaphor, “The best defense is a good offense.” (That’s not us, it’s Napoleon. Who, aside from being a dictator, was a pretty successful strategist.)

So, why do so few companies adopt a proactive strategy?

From what we’ve found, this is largely due to the fact that they lack the necessary data. Or, more precisely, they have LOADS of data, but it’s not refined enough to provide pivotal insights. Unfortunately, a lot of the current metrics available (Klout or PeerIndex scores, friends/follower counts on various networks, etc.) only give you a basic sense of how popular an individual is, how often they engage, and some of the things they talk about. Some of this data can absolutely be useful for reacting to unique mentions and comments as they come down the social pipeline…

“@kimkardashian mentioned us. Let’s freak out!”

“@scobleizer talked about a ‘boarding pass’ – let’s pitch him about our travel backpack.”

… but responding to lightning strikes and keyword filters aren’t really long-term proactive strategies.

In order to truly be effective when engaging online, you need to lead the conversations and provide benefit beyond what flies across your social dashboard. You need to engage with individuals in the outlets and channels they’re already using. Why? Because we believe the Cluetrain Manifesto came true: networked markets exist for everything. Of course you have to engage in those conversations. But there’s also a bigger opportunity here.

There are people online who have a terrific understanding of your market and – more importantly – who already have the attention of the people you’re trying to reach: the potential customers for your products. Shouldn’t you be asking these contextual Influencers what they want in your next product? Wouldn’t it make sense to get their input on messaging before you launch that $2M ad campaign?

Until now, it’s been very difficult to be proactive online. Social media is fragmented across the web, the content is largely unstructured, and there hasn’t been an effective solution for discovering the influencers around any search – and definitely not at Google-speed. That’s what we’re aiming to solve at Spot Influence. Through data mining, lots of Big Data tech, and our own fancy algorithms and analytics, we’re able to discover what everybody cares about, where they’re posting content, and who they listen to on specific topics.

At the end of the day, we want to help people be strategic with how they approach social media. Our data makes it possible for every company to do just that. The future of online engagement is being proactive.

The Death of Demographics

Posted on by
Reply

A few months ago, our Co-Founder Dave Angulo did a presentation at the Defrag conference regarding “The Death of Demographics” [See SlideShare deck at bottom]. Recently, I was chatting with the president of an advertising agency, and he mentioned a similar theme. When describing our data, he felt that our technology could revolutionize the industry by providing access to people data at scale – eliminating the need for generalizations of large audiences. We’ve gotten a really positive response from people regarding this topic, so I think it’s time for a blog post…

Traditionally, demographic analysis has involved dealing with generalized abstractions of people, and then picking a subset to ask detailed questions in order to understand them better. A great example of this is what Time magazine calls, The Smartphone Mom. According to Time, “Web firms are collecting personal information about moms, including what times of the day they’re logged on, if they are connected from home or on the road, and how often browsing turns into a purchase. Firms such as Procter & Gamble, Walt Disney, Comcast and AT&T… want to use that kind of data to tailor ads to that demographic.”

Demographic data can be very useful for marketers, but it’s limited in the sense that the insights are only generalizations. For example: “52% of Moms said they use their smartphone within 5 minutes of waking up” (Life360). As a business, don’t you want to know who those 52% actually are? Specifically, what are their interests & where can you engage with them online? Are there influencers within that specific group of Moms?

These signals have existed online for some time, but up until now, technology has prevented them from being analyzed at scale. Generalizations were made because understanding the interests of millions of people, at the granular level of each individual, was impossible. There was simply too much detailed data to provide business with actionable insights. With the technology we’ve developed, we’re about to change that…

By organizing the web around people, Spot Influence is able to determine who the influencers are for any subject/search online, where these individuals post public content (Twitter, Blogs, LinkedIn, etc.), and what their interests & influential topics are. This data has the potential to drastically change how companies market their product & gain insights into their actual audience. Generalizations no longer need to be made. We can’t wait to see the impact this technology has on Marketing!

Check out our presentation from Defrag 2011 below, and let us know your thoughts on this post.

Death of Demographics

 

The French Internvasion: News from the Battlefront

Posted on by
Reply

Hi, it’s Rémi, Spot Influence’s new French intern. Last time around I told you why I’m here. Now it’s time to tell you how I’ve been doing so far.

Adapting to Boulder has proven surprisingly easy. Sure, the weather is all over the place (40F differential in 12 hours, come on!) and the city is very different from what I was used to (i.e., it’s really roomy and car friendly), but the omnipresence of bikes is comforting. And the American TV series are the same on this side of the pond, so I already feel at home here.

Another thing I really enjoy about Boulder is its vibrant Tech community. In the few short weeks I’ve been here, I’ve been to one Boulder Ignite event (amazing), one Boulder Beta event (awesome), and the TechStars party – which was tons of fun. Such a concentration of talented geeks is something to be treasured because it creates daily amazing opportunities. Spot Influence was born out of one of those community events!

And I’ve even had California wine which was pretty good, despite all I’ve heard about it at home (California wine, pfah! it’s not French…). I suppose the hardest thing to bear is that silly 400% tax on Roquefort (the mother of all blue-like cheeses), but I’m pretty confident I’ll eventually get over it.

On the business side, things are going great here at Spot Influence. After several weeks devoted to getting up to speed on all the top-notch tools we use, I’ve started diving into the codebase to get a grip on how we solve the complex problem of calculating contextual influence at scale, and how exactly I will be able to help here. Obviously I can’t spill any corporate secrets, but I guess I’m allowed to say this: what we’re trying to do is HARD! I’m amazed at the amount of work we’ve already accomplished and I really look forward to adding my small contribution to the mix. There’s no reason to be surprised about this; the people who work here are just plain smart and they’ve had a bit of time to come up with an answer. Plus my colleagues are all super nice, so I’m pretty happy where I am, thank you very much. All I have to do now is to measure up. That’s a bit scary, but I like a challenge!

And now for what you’ve all been waiting for (hence the cryptic title of this post), news from the not-so-hidden war the French are waging to conquer Boulder. I’ve already arranged for several friends to come pay me a “visit”. Let me warn you, most of them will be armed with skis and poles. I’ve also discovered that there is another Rémi (typically French name) in Boulder. Our network is growing and pretty soon, we’ll be unstoppable. The soft-power conquest has already begun: at the office, I’m slowly converting everyone to French wine even as they believe they’re converting me to American beer.

Marketing & Social Data in 2012

Posted on by
Reply

Recently, there have been a lot of 2012 prediction posts talking about marketing & social media trends for 2012 and making recommendations for how folks should alter their current strategies.

The following ones impressed us in particular…

 

Additionally, we wanted to make some of our own 2012 predictions regarding contextual influence & social data, and how we think they will be used by marketers and businesses in the new year.

Engaging with the “right people” will become a key marketing strategy. 

Marketers will spend less time questioning the benefit of engaging in social media and more time actually engaging in it – especially via data-driven strategies tied to ROI. An important piece of this strategy will be determining who to engage with. Contextual Influencer data will become critical to effective Word-of-Mouth campaigns and will prove it’s value through increased spread of messages and engagement within relevant communities.

“Social Media Marketing” will become “Marketing.”

Having a deep understanding of Social will become a core competency for marketing professionals and agencies. As social data becomes increasingly valuable, a greater portion of the overall marketing budget will be allocated to data mining, analysis, and strategic engagement. In addition, social signals and data will increasingly be integrated into other areas of marketing.

Social Data will become fully integrated into the CRM & Sales process.

As social profiles become easier to acquire and integrate, CRM & Sales will increase their usage of social data and channels. Strategies will emerge regarding optimizing the flow of information online and that will have a clear impact on overal sales and costs of customer acquisition.

Marketers will demand distilled, actionable data. Big Data will become small.

Getting lots of data is easy, making sense of that data is hard. As access to “Big Social Data” increases, business will demand services and tools which can provide distilled, actionable insights and increased conversion rates.

What are your predictions for 2012?

SoftLayer Technology Partners Marketplace

Posted on by
Reply

Spot Influence was recently featured in SoftLayer’s Technology Partners Marketplace!

SoftLayer’s Technology Partners Marketplace is a showcase of current SoftLayer Software-as-a-Service (SaaS) and Independent Software Vendor (ISV) customers’ products and technologies. The Marketplace is designed to connect these customers with the rest of SoftLayer’s community of over 26,000 customers.

We’re very excited to be a SoftLayer customer and a part of the Technology Partners Marketplace. We’re looking forward to working with them, and building a great business relationship!

Be sure to check out our guest blog post and company profile for the Marketplace.

Additionally, here’s the video interview we did:

The French are Coming!

Posted on by
Reply

Since the end of November, there’s been a strange and slightly frightening addition in the Spot Influence office: the French Intern. Mostly he just sits there behind his brand-new computer and tries to catch up with all that he needs to learn, but who knows what he might really be up to?

Oh, that guy… well, it’s me.

So, apart from being French and being an Intern at Spot Influence, who am I and what the hell am I doing here? My name is Rémi Leblond. I’m 23 and I graduated from a French engineering school a year ago with majors in Applied Mathematics (in French, that actually means “not-extremely-theoretical Maths”) and Economics, and a minor in Computer Science. I then enrolled in a government program to become a senior staffer in the industry ministry. This is my second year in that program and it basically consists of a one-year internship abroad. That it be abroad is just about the only limitation to the internship – which means that I might have chosen to do anything, from selling mobile phones in Timbuktu to wasting my youth at the Goldman Sachs.

So why intern at Spot Influence? Several reasons, really, which I will now give in no particular order.

I guess the first reason is that I wanted to work in a start-up company. I’ve had an experience in a big corporate firm in France and I wanted a different angle this time around. So when my cousin who lives in Denver told me “hey, you should probably check out start-ups in Boulder, there are some pretty cool companies there, plus the ski is amazing”, I took the bait instantly.

I began researching the subject and I very soon realized that although Boulder might not have the same reputation as Palo Alto, it is a very rich and dynamic hub for technology startups. I quickly found the TechStars website and its handy description of every startup that had been through the program in the last few years.

Among that list, I found Spot Influence. From the get-go I was taken in by the project. Influence is an extremely hot topic nowadays for marketing firms, PR firms, campaign managers, etc., and no one seems to have yet found a satisfying solution to that problem. And as I watched Dave and Rich’s video, I realized that Spot Influence’s approach of contextualizing influence is what really sets it apart from its competitors. Don’t get me wrong, absolute influence isn’t worthless, it’s the subject of innumerable “100 Most Influential People” articles every year, in which we discover to our dismay that the Pope has been overtaken by Kim Kardashian.

But what really matters is whether people are influential in given fields. Pampers isn’t really interested in the fact that Ashton Kutcher has 20 zillion followers on Twitter. What they need to know is, “Who are the people driving the internet community that talks about diapers?” – so they can reach out to them and run efficient marketing campaigns.

So my first thought was “Wow, what these guys are trying to accomplish is amazing.” And my second was, “Wait a minute, isn’t this exactly the kind of company which might be interested in my profile? I *did* major in applied mathematics with a focus on graph theory after all. I might actually bring some additional value to the team.”(if you’re counting, these were reasons two and three).

I was hooked; I plucked up my courage, updated my resume and sent it along with a cover letter to *hello@spotinfluence.com* and I started praying for an answer. I didn’t have to wait long: three days later, Dave answered that they were interested in talking to me. Interviews with the team followed and it all went very fast from there. And so the fourth reason was that the people at Spot Influence had a very open mind when it came to who they wanted to work with, and that they weren’t afraid of putting in the necessary administrative legwork to make it so I could come and help out. Realizing that made me even more impatient to work at this company.

For all these reasons, I chose Spot Influence. Having been here three weeks now, I can assure you that I was not wrong. But this post is already very long, so I’ll wait until the next one to tell you my first impressions of Boulder, working at Spot Influence, and what I’m looking forward to in the near future.

Stay tuned for more episodes of the French invasion.

Influence Without Context is Popularity

Posted on by
4

In the last few months, there’s been a lot of online discussion regarding influence metrics. As we get closer to launching our public Beta for Spot Influence, we think it’s a great time to talk about our view regarding influence and why it needs to be contextual.

Right now, the majority of influence metrics are focused on measuring a person’s GENERAL ability to engage an audience online. They measure the # of Friends, Followers, RTs, Mentions, Shares, Likes, etc., and boil that down into a single numeric influence score that gives insight into how good they are at spreading messages online. Those single-number scores can be useful. For example: If you sell hammers and Kim Kardashian mentions you, congratulations! You’ll probably sell more hammers this week.

However, Kim mentioning your brand is a lightning strike, not a repeatable strategy. As a hammer manufacturer, it doesn’t make sense to reach out to Kim. Why not? Because, although she’s popular, Kim doesn’t have the attention of the people who actually care about hammers and might buy your products. Kim isn’t talking to your potential customers.

This example demonstrates that there’s a difference between finding popular people who might use relevant keywords (or even have them in their bio), and identifying the people who actually drive the conversations that matter to you and your brand.

If you’ve been active in social media for a while, you know at least some of the people who are important to pay attention to and engage with. You’ve also realized that this approach takes a lot of time, is incredibly subjective, and you’re probably not tracking everyone you should be.

Fortunately, relevant contextual influencers - who matter to you and your brand – actually exist. The challenge has been finding them.

We’ve been working hard on this problem at Spot Influence. While other companies are great at building tools, running marketing campaigns, and managing communities, we’re all about data. Our goal is to know as much as possible about the people creating publicly-available content online: where they publish, what they write about, and how effective they are at engaging others who also care about those things.

Stayed tuned. In the coming months, we’ll be posting details about our API and the tech we’ve developed to uncover contextual influence at scale.

Rackspace Startup Program

Posted on by
1

The TechStars 2010 program was an incredible experience for us, and now that we’re TechStars’ alumni, we’re incredibly fortunate to have access to some pretty awesome perks. One that we’re particular thankful for is the Rackspace Startup Program.

The program provides hosting and cloud computing resources to young companies, and they’ve partnered with startup accelerators and incubators nationwide to provide their services, including: TechStars, 500 Startups, Y Combinator, and General Assembly.

Here’s a great video of Rob La Gesse (Rackspace’s Chief Disruption Officer) describing what the program is all about. If you are interested in the program for your startup, you can sign up here to learn more!