Algorithmic Campaigning

Categories: Uncategorized
Comments Off

Had a lively discussion today about algorithms and the Presidential campaign. I was being asked to attend a party meeting and see how I thought one party could get their candidates elected. My answer? Algorithmic campaigning.

First up? The candidate… and look no further than Rick Santorum swearing at a reporter for being misunderstood. Who’s fault was that? Santorum’s. Can it be prevented? Absolutely. It just takes a little math!

IDEA Algorithms can read and analyze every piece of news reported on a presidential campaign, including comments by citizens. Hours after a speech in Dallas, every word uttered by anyone can be captured and analyzed to see if the candidate was misrepresented, misquoted, misunderstood. But here’s the kicker. These algorithms help write the speeches. What resonates in this city versus another campaign stop? Where does the candidate have to tread lightly? What language does the candidate need to use to drive a point home. An hour after the speech, we will know how the candidate did and what to change for next time… Every critique, position, comment is cataloged and classified. This takes knowing your audience to a whole new level!

But it doesn’t have to be after a speech. If the candidate is coming to New Orleans, the speech process can be informed by the algorithms before the plane touches down. Newspapers can be analyzed, columnists, bloggers, you name it. We are not asking a candidate to change their position. We are asking the candidate to focus on certain positions that resonate and will likely get support by the media in that city. We are asking the candidate to stress certain positions over others. These algorithms can even help inform who the candidate should meet with and grant interviews to, both friendly and unfriendly. The math determines the influencers in each market.

Not only does this work in Dallas and every other campaign stop, it works on the national level. Heck, it will work with Wolf Blitzer. Before a candidate sits for an interview, a full dossier is generated by the algorithms. Candidate knows when Wolf is going to interrupt an answer before Wolf does!


Idea #2. Fundraise using resumes. Download resumes and analyze them down to the sentence. Send very specialized campaign emails to each thin slice of the electorate. You work in the healthcare field, you get specific emails. Work on Wall Street, you get different emails. It’s about customizing the message down to the voter’s specific areas of interest. You were a Boy Scout, here’s one message. You went to BYU, here’s your message. You have a Masters, same thing. And the analytic value will be huge. I know every article the voter opened. I know if you forwarded my campaign email. You never click on an article that criticizes the other candidate… no problem, you will never see those articles again.

Algorithms help manage the candidate message for maximum impact. And algorithms help with highly-tuned fundraising. After all, not an email is sent that doesn’t ask for money. But at least this email speaks directly to the voter.

Communities of Experts

Categories: Uncategorized
Comments Off

Before Coopers & Lybrand had email, we had “Calling All C&L.” It was a message board for any employee to post a question and get answers from around the firm. This was 1992. In 20 years, I have never witnessed a piece of networking technology as successful as Calling All C&L. It was extraordinary.

What happened? The cost of entry went to zero. We commoditized communication. Made it liquid. Facebook is a mix of posts about a parent passing away, someone needing “crystal shards” for an online game and a post with the single word, “bedtime.”

IDEA  Raise the cost of entry. Qualify each user for a spot in the social network. Don’t focus on users’ collecting thousands of friends. Instead have the network (algorithms) make the connections. Users are qualified, then invited to join the network. From there, each is matched to other users’ profiles to build areas of commonality. It’s Facebook, but with no button to add friends. Experts are added to your circles automatically.

The key to all this? Resumes. Analyze each resume for expertise, schooling, years experience, current position, etc. Group experts together based on resume content. Allow any user to recommend someone for the network, but the cost of entry? You submit your resume, have it analyzed and be automatically assigned to groups.

Sales: Information Flows are Key

Categories: Uncategorized
Comments Off

Let’s put together a 3 inch binder on General Mills and have an all day targeting session. Net effect. None. Why? What’s in today’s binder will not help us sell a thing tomorrow.

Revenue from new clients requires a campaign approach. Sales professionals need to stay front of mind and wait for their opening. It may be the current provider makes a mistake or the target company needs a second opinion on something important. You need a reason to contact the target. You need a conversation starter. We call this a “touchpoint.”

IDEA Go for the long tail of information. Everyone can do a Google News search or pick up the Wall Street Journal, but what if computers could scour the entire internet, everything from trade journals to Twitter. There are plenty of clipping services out there, but they either deliver the most popular articles or all the articles sorted by date.

We are talking about a smart algorithm that determines the most opportunistic General Mills events each day, from a sales point-of-view. Sales professionals find this approach not only starts the conversation, but the targeted executive often asks for a copy of the article, since it’s the first s/he has heard of it.

6 Degrees of Separation

Categories: Uncategorized
Comments Off

About the time Anthem and Wellpoint were merging, we were trying to win a significant engagement with the merged firm. We knew Wellpoint very well, but had no relationships at Anthem.

Our clients included a company on the west coast that made RVs. Turns out one of the board members at Anthem was also a board member at this RV company. One call later and we had both intelligence and a way in to this board member.

Another Anthem executive used to be CEO of a Boston-based company. They, too, were a client. A call to our Boston team revealed we had along-standing relationship and an employee who attended church with the Anthem executive.

IDEA Using named entity recognition, executive bios can be parsed for past employers, board memberships, even non-profit affiliations. This information is loaded into databases along with our current client list and ally our team member assigned to those clients. Searching on an organization name instantly brings up all zero, one and two degree connections into the targeted company.

Walked into Microcenter the other day. They asked if I would like my receipt emailed, printed or both. Did I really want to give Microcenter my email address, so I can get spammed? And what’s the value of the receipt in the first place. Mine sit in a shoe box on top of the refrigerator.

Retailers are desperate to tap your buying habits and sell you more stuff. While I’ve been carrying a Kroger Loyalty card for years, I’ve never once received a personalized email featuring products I usually purchase.

IDEA Strawberry pop-tarts sell before a Hurricane. Diapers and beer are a staple of Friday nights. Nacho cheese dip can be found on the chip aisle. These are examples of obvious relationships. But hidden inside the petabytes of receipt data are strong correlations between highly disparate products. These are products that the masses (crowd sourced) think go hand-in-hand, but perhaps you haven’t made the connection yet.

Today Kroger pitches the same sale on diapers to everyone, even single guys. Next up? If I buy Huggies, I will probably get a coupon for Pampers. But I already like Huggies. The real solution is to connect specific diaper sizes to other products in the store. One day, I may buy Huggies in a toddler size and get a coupon attached to my receipt for a family sized frozen lasagna. The key is not to give me deals on stuff I already buy or competitor products. It’s to lead me to purchase something I would have never considered I needed.

56 percent of students who enroll in a four-year college earn a bachelor’s degree. More than half of first-year students are simply underprepared for college-level work.

College admissions officers are reviewing grades, SAT scores, teacher recommendations and essays. Admissions is asking whether the student deserves to be at that institution. Perhaps, they should be asking a different question… Will the student make it to graduation?

IDEA Hidden deep in the details of a college application are correlations that determine the likelihood a Freshman will make it to senior year. What does the student’s father do for a living? Did Mom go to a similar school? What was the size of the high school? The size of the town? What grades did the students get in certain electives? Is the student the first of the siblings to attend college?

Not only are the answers in these applications, they are provided by the graduating seniors. I can see admissions doing exit interviews with seniors and changing the incoming application process. One day, we may ask on a college application how many nights a week the family has a meal around the dinner table. This could be a measure of student independence as much as it is about family support.

Doctors don’t like electronic health records (EHR), according to a recent article in MIT Technology Review.

Patient presents with lung cancer — undifferentiated tumor in her mediastinum. Doctors don’t want to operate because there is a 15% risk of death with surgery. Within a month, patient dies, not of lung cancer, but suffocation. Turns out tumor was pressing on diaphragm and making it hard to breathe. Biggest fear during entire ordeal… don’t let me suffocate to death. Happiest during the process? Sitting with other patients discussing types of wigs.

IDEA The answer to EHR lies in the support group, not the doctor’s office. While every cancer is unique, there are non-obvious relationships that bind all these patients together. Imagine computers analyzing thousands of lung cancer patients and thin slicing each into very small groups. These groups belong to slightly larger groups, etc. In other words, patients form classes within a taxonomy.

Each new patient gets classified and algorithms use this information to find patients with very similar (EHR) records. The computer may immediately recommend a brain scan to see if the tumor has metastasized. Algorithms know that in this patient’s disease classification, X% of patients also have brain tumors. Algorithms would also tell surgeons that a 15% mortality rate is acceptable. Get as much of the tumor as possible to provide relief to the diaphragm. The algorithms are self-learning and get smarter as more patients are added to the database.

Textbooks in Education

Categories: Uncategorized
Comments Off

Primary and secondary education are not going very well. We have examined every possible cause except one, the textbook.

Put in a Little House on the Prairie DVD and I swear Laura Ingalls Wilder could be attending my daughter’s school. We have gone from horse and buggy to cars to planes. Computers replaced calculators, which replaced slide rules. Yet, we are still using antiquated technology (textbooks) in schools. And, no, digitizing that same textbook for an iPad is not going to make a difference.

IDEA Replace the textbook with blueprints. If architects can build skyscrapers with them, I think we can teach today’s youth. Imagine a blueprint on science. Where does matter fit? How does it relate to chemistry? Does chemistry lead to cells? Blueprints connect the dots. Students no longer memorize, but instead learn spatially. Close your eyes and picture the answer.

And each chapter or theme in the old textbook could be a blueprint. Start with the science primer, move to a blueprint on motion/forces. From there move toward matter and onto chemistry. Everything ties together. And the blueprints cost a couple dollars to print. Each student is given a set to take home. Can you imagine parents reactions when their child shows up with science blueprints. Tax dollars very well spent!


Sales Force allocation baffles most companies. At the end of the day, best people work the largest clients. I feel for the new employee. S/he often gets the worst clients dumped at his/her feet.

IDEA Client spend is not the most important criteria. Take someone who lives in a $10M estate in Beverly Hills. This may be a $50 million client for Merrill Lynch, but do they realize just down the road is another client who only invests $1M with the firm? In other words, if I put my best sales professional on client two, is my potential upside ~$49M, assuming they live in similar homes?

Doesn’t take a rocket scientist to see using real estate value is a great idea. But what if I told you there are more than 50 factors that can be used to evaluate the “upside” of an opportunistic client? Everything from bio information to things like commuting distance.

I must have signed up for Quora, because now I get a weekly digest. Interestingly, they call it “my” digest?

But it’s not personalized in the least. I have no interest in Craigslist’s genesis. If I did, I would go to Wikipedia. Same with favorite parable. I suspect Quora sent 1MM of these weekly digests and we all pretty much got the same material. Why?

Perhaps if I was a heavy user, an algorithm would have learned my interests? But I am not a heavy user. And this digest isn’t helping.

IDEA Don’t source interests from my activity on your (Quora) web site. This requires too much of a training set. If I post 1000 messages, I will get a better digest (maybe). But it’s the first 10 messages I post where you are trying to convert me into a regular user.

Ask me where I work, then index my web site. Ask me my position and index job postings with similar titles. Ask for my LinkedIn and Twitter URLs and index those too. I provide Quora metadata about myself. One layer below that metadata is enough data to classify me into a thin slice… a starting point.

For example, let’s say my title is marketing manager and I work at a Big Four accounting firm. Quora spiders every job posting for marketing at an accounting firm, analyzing the resulting text for themes (popular keywords/phrases). Quora then uses these same themes to search Quora postings each week. The intersection of the two? My personalized newsletter.