Friday, March 20, 2015

Andrew Abbott on the Nonlinear Nature of Research

Andrew Abbott is a sociologist at the University of Chicago who specializes in, well, specialized knowledge, the disciplines and also research as a practice.  He has spent his career doing research, writing about it (very meta), and studying how students and professors develop ideas.

His latest book, Digital Paper, is a how-to book on doing research projects.

What is a nonlinear process?

Spoiler Alert! Abbott is going to describe a research project as a nonlinear process, but does that qualify as news?  Here in the 21st century creativity and right-brain-edness have become not only accepted, but seen as key to success in business even outside the creative bastions of marketing and advertising where, I'm told, workers typically put their feet up all day and doodle.

A linear process is a series of left-brained, logic steps.  Like cooking a meal from a recipe.  Or the many scripted things that we do in life.  When you order food from a waiter in a restaurant, that's a linear process.  When the waiter goes back to the kitchen and hands off the order to the chef who follows a recipe to cook the food, that too is a linear process.  The behaviors that young children learn are all linear because nonlinear is too advanced.

And nonlinear is also hard to describe.  I can give you directions to Greenwich Village or a recipe for Coq au Vin, but how do I tell you how to produce a research paper that says something new and interesting and have it come out as good as the food.

Curtain up, enter stage left, Andrew Abbott.

Research Projects are Nonlinear

When I searched for "How to Do a Research Project", the first hit was an article on how to do an A+ paper.  Answer?  7 steps: choose topic, find information, state your thesis, make an outline, organize notes, write the first draft, revise, and finally, type the final draft.  Finally.  The cat is out of the bag.

But wait...Abbott writes that when you do research,
You don’t start with a general question, focus that into particular questions, then specify the data you need, gather the data, analyze it, and finally write up the result...Quite the contrary, you will be doing many different kinds of things at once. Only at write-up time will you cast the project into the classical rhetorical form: general questions leading to specific questions leading to analysis and finally to conclusions.
He's critiquing any kind of linear step model, especially one that relies on getting the plan or design of the research right in the beginning.  Instead the plan emerges as you proceed.

Nonlinearity is especially descriptive of the browsing and searching process.  The researcher needs to take advantage of randomness.  Furthermore, the author writes, "you surrender to nonlinearity."  This language suggests that people--especially people with a deadline--often have a resistance to the chaos of the back and forth and they have to give that up in order to get the job done.

Construct Your Work from Infinitude

Research is not about finding things and putting them together. That might be linear.  Even though research involves a massive but finite set of materials in libraries and on the web, the number of questions that can be asked about those materials is infinite and the number of combinations of those materials is infinite as well.  So research involves constructing an answer to the puzzle that was posed

The Process Revealed

What were steps in the primitive linear description been metamorphosed into seven tasks that are orchestrated and dynamically sequenced based upon the current needs of the project.  Those tasks are:
  • Compiling Bibliography
  • Scanning and searching
  • Reading
  • Maintaining files (organization of assets and artifacts)
  • Analyzing material (minianalyses)
  • Writing

The Temporality of Nonlinear Research

A process is something that unfolds in time, so merely breaking tasks out of their lockstep martial format isn't very helpful.  What is needed is a kind of temporal map and therein lies the genius of the book.

Abbott presents a big-picture timeline of the seven tasks across five phases of a research project: preliminary, midphases 1,2, 3, and endphase.

During each phase, the researcher is performing the same tasks but the relative times and the specifics of task execution change.  More time is spent writing toward the end and reading in the beginning is more background-oriented. The book goes into the subtleties and changing patterns of these tasks at different points.  Here the book shines and a summary falls short because it can't capture the back and forth of the process.

In the preliminary phase, the researcher starts with a general interest and formulates it into a more specific puzzle.  Then he browses and scans sources in libraries and on the web to produce an initial bibliography.  The researcher also does background reading of whole texts.  At some point in the process, the design document--the plan of the research--feels stable.  That's how you know you're in midphase.  Now the researcher is working with the materials he's collected into a bibliography, but in the process leads arise that inject nonlinarity in the form of fishing expeditions.  Along the way, minianalyses and mini-writeup are done to absorb and process the research data. The researcher is in the endphase when he's ready to do most of the writing.  Linearity returns.  The fragments of writing are assembled and tweaked and transitions are added.  Text is revised and gaps are addressed.  The researcher arrives at a draft.

There Will Be Fishing Expeditions

Whenever I've encountered the term fishing expedition in text or speech, it's always been negative: someone is wasting time following a lead that is doubtful on the surface.  But for Abbott, a fishing expedition is a normal and regular process in research.  Researchers have to take these risks.

In chapter two, where Abbott presents a case study of his own research, a paper on the history and habits of professional researchers, he decides to read the title of every thesis ever done at the University of Chicago Graduate Library School.  His thinking going in was that 10% of the titles would be useful and a few would be absolutely central to his work. So, against the advice of most books on research, he brute-forced his way through thousands of theses.

Abbott says that to be successful in a fishing expedition, you need to be able to recognize that it's a fish when it lands in the nest, but you care less about the type of fish.  Abbott caught 600 fish in his net, scanned the 600 dissertations and, to the librarian's chagrin, checked out 100 useful titles.

Did You Know? Monkeys are Nonlinear

Monkeys and gibbons brachiate as they travel through the forest using their arms to swing from branch to branch.

Abbott explains that the locomotion of research also involves brachiation.  The researcher moves back and forth in time (older and newer texts) and place (texts written in different places) and landscape (different types of text) to follow the clues.

For example, a researcher will find a source, perhaps disregard the text itself, but after scanning the indexes and bibliography, discover concepts and references, and then swing to another source, repeating the pattern and moving through vast physical and digital information cosmos to find the right information.

A Researcher is a Human Virus Checker

Abbott writes that browsing is going on at all levels at all times.  The researcher is like a human virus checker, scanning the information landscape with peripheral vision looking for finds.

Nerdbar: Software Development is Nonlinear


When I first saw Abbott's timeline, I was reminded of the Rational Unified Process (RUP), a software development methodology which was developed in the 90's (under another name) to deal with the fact that too many software projects were never getting completed or finishing late and grossly over budget.

RUP may have been the first methodology to incorporate non-linear elements into the development process and has parallels to Abbott's research process.  RUP has activities: requirements definition, design, implementation, etc and those activities are iterated across four phases.  Like researchers, software developers are designing at all phases, but doing more design in the beginning.

The following diagram vividly illustrates the proportion of time spent on the core activities at different times.

Organization of Project Artifacts = Thinking

Regardless of how you organize your files, physically or digitally, organization is important--for the reason that the researcher is generating a lot of content in the form of notes, writeups and copies of sources and that content will need to be found and re-found as the project progresses.  So organization is important for that reason, but Abbott makes a second point:
Doing the filing is thus a central part of the intellectual work of a project.
In other words, organizing is a type of thinking that contributes to the content of the project.  So findability is not a sufficient criterion for an organizational style.  Your organization system must help you analyse what you're organizing.

Abbott provides a list of artifacts that the researcher generates.  He talk about the folders you need; I've translated to artifacts.
  • correspondence
  • task list and task log
  • current and archived design documents
  • writeups,  minianalyses, and notes
  • primary sources (like interviews, collected documents)
Abbott cautions that if all you do is tag documents, you're refusing to think.  He seems to equate thinking with putting things into folders.  Later, he clarifies his point by asserting that these organizational acts involve creating a controlled vocabulary of the terms and concepts that will be the key concepts of the work.  Perhaps Abbott would be OK with tagging if the tags were thoughtfully chosen and reflect analytic thinking about the research topic.

He puts the point in another way: can think of your final written product as a thoughtful and even authoritative index to a certain set of materials. It is an index from a particular point of view— yours. And the claim you make by writing your final text is that you have a particularly wonderful index to your materials.
By implication, if you're using a tagging system, it should reflect your controlled vocabulary.  Then tagging would be a key part of the thinking process. Tags and names for artifacts are an invention.

Abbott writes,
It forces you to reduce ambiguities in your thinking (“ did I think that article was about anxiety or about fear?”), making the judgments and inferences that gradually constitute your analysis.
 Just like the human virus checker scanning for leads, organization and re-organization goes on throughout all phases of the project. It's part of the main work.

Abbott suggests a rule of thumb for categorization which he calls the six item rule:  don't put more than six items in one category.  Again, a similar justification.  Creating the sub-categories forces more analysis.

A Concluding Butterfly Postscript

Nonlinear has another meaning: that a small change can have a large and disproportionate impact: a butterfly flapping its wings in Taiwan causes an earthquake in California.  Likewise, a discovery of a small piece of information may trigger a redesign of your project.

May the butterfly flap its wings early on in your work!

(This post was written with a new web app, AirStory, that I've been developing with Joanna Wiebe.  If you're interested in trying it out, email me at jim underscore briggs at athenz dom com.)

Thursday, December 27, 2012

On Remarkable Products: Seth Godin's Advice in Purple Cow

I guess I'm in the late majority because I just got around to reading Seth Godin's Purple Cow: Transform Your Business By Being Remarkable.  By now, the ideas in the book are less novel, but just as, or more, urgent.  And the writing is colorful and logical: a rare combination.  I expected another derivative marketing book; boy was I surprised.

The setup of the book reminds me of the setup in Daniel Pink's A Whole New Mind:  Why Right Brainers Will Rule The Future :  Consumers are less needy, more satisfied, have more choices, and less time to pay attention to ads or marketing about new products.  The old marketing and advertising strategies don't work anymore.

So what do you do?  Godin: Take your marketing budget and invest it in making your product remarkable.  But not remarkable to anyone.  Remarkable to the innovators and early adopters in your market who will be so impressed that they will spread the word for you about your remarkability.  These sneezers will sneeze your remarkability to the rest of the market.  To sum it up in one phrase:

be remarkable to market innovators.

The book replete with examples of remarkable products and some corollaries of the above thesis.

Corollary #1.  It's risky to be safe, or the inverse: safety is risky.  Now as an absolute, this is not true, but it points to the idea that getting to the remarkable requires taking some risks to do or find/create that remarkable rare bird of a product.  The desire to play it safe and avoid risks, especially among successful companies makes it harder to achieve the remarkable.

Corollary #2.  When you go out on the edge where the innovators are, you'll develop something that the market majority may not understand.   So, very bad can be a sign of good.  (This has to be balanced with other considerations such as can I make enough money from innovators before I cross the chasm to the marjority in the market.

Qualification. Its possible that in some markets, the innovators may not be merely bridges to the market majority but cash cows  for your purple cow.  His example: online banks that derive most of their deposits from the customers that use the online banking features.  Pearl Jam is an example because they made a big business out of selling their live concerts to the innovators.  It's a great example of growing the customers you already have.

Corollary #3.  Boring leads to failure.

Godin argues that targeting innovators is so important that you should pick markets that have innovators that can serve as sneezers.

Below are some examples of remarkable products.  Some of these companies may no longer seem remarkable because remarkability has an expiration date which may be reached when a product becomes popular and common and the company fails to continuously innovate.
In pursuit of the purple cow, here are some questions to ask yourself:
  • Which markets have the most innovative sneezers?
  • In a given market, what customers are most likely to be innovators?  A sneezers?
  • How would you Bronnify, Kiwi,-ize, Aeronize, South Park-ize your product?
  • If you won the lottery and gave it all to a team of designers and usability experts to design your next product, what would it look like when they were done?
  • What slogan would you write to appeal to the sneezers?
  • Who in your industry has a great tracking of launching remarkable products?  How would you recruit them or copy them?
  • Pick a remarkable company in a dull industry and do what they did.
  • What are you not doing as a result of fear?
  • How can you out do a company in your market that is already on the edge?

Thursday, December 20, 2012

The New Entrepreneurial Methodology

A new entrepreneurial methodology is emerging under name of lean startup (with customer discovery).  The spread of the methodology coincides with a worldwide spike of interest in and practice of entrepreneurship.

The methodology can be summarized as follows:
  1. Define a business model (the plan)
  2. From day one, develop personal relationships with customers.
  3. Run a customer-centered experiment to test your plan.
  4. Based upon what you learned in the experiment, revise your plan and your product and design and run more experiments, repeating the process until you find a plan that works.
Or to summarize it even more concisely, in six words:
  1. Business model
  2. Customer relationships
  3. Experiments
  4. Rapid cycle time
The uncertainty premise.   This methodology rests on one critical assumption:  that starting up a company to create something new or even something new to you is a highly uncertain venture.  Many startups today in the web and mobile space are doing just that.  On the other hand, if you've managed a beauty parlor for years and you want to start your own beauty parlor, there's less risk.

Ironically, entrepreneurs seem genetically programmed to underestimate their risks and to act as if they didn't exist.  This is bad for entrepreneurs but good for society which thrives on many people trying and failing to create new things.

Back to the methodology...

1.  Business model.  A business model is, while having a fancy name, a simple thing.  Its a simple representation of your plan that answers the questions:  
  • What am I offering (value proposition)?  
  • What are the customer groups that will buy my offer (segments)?  
  • Will customers actually buy what I'm offering (problem-solution fit)?  
  • Will enough customers buy from me (market size, product-market fit)?  
  • How will I get, keep, and grow customers (customer relationship)?  
  • How will I make money and will it cover my costs (revenue streams and cost structure)?  
  • What resources do I need to make it all happen (key resources)?  
  • What are my most risky assumptions (key assumptions)?  
A business model can be summarized in one or two pages.  It's simpler than a business plan, more useful and represents a complete picture of the business along with its key risks.

A business model is also a set of guesses.  The method helps you turn guesses into facts.

Lastly, a business model is a logical expression articulation of an entrepreneur's vision which is the creative powerhouse and engine behind the creation of new things in the world.

2.  Early and personal customer relationships.  Its important to develop personal relationships with customers early on even if your starting a business that doesn't require much ongoing customer contact.  Customers provide an early warning system for flaws in your business model.  They give you deep insight, often unintentionally, and that insight  leads to changes in your model, small and large.

3.  Rapid cycle time.  You won't succeed with your first business model.  Perhaps not with your second, third, or tenth.  This is a foundational fact.  

Let's engage in a thought experiment.  Imagine you could look into the future and know that your tenth plan is the one that will work.  You can't jump to ten because you're going through a learning process where one plan leads to the next.  At your current rate of learning, you will run out of time and money before you get to ten.  So the solution is to speed up the learning process.  The more quickly you can loop through a plan-test-revise cycle, the better your odds of success.  This is also called iteration.

4.  Experiments.  Experiments are what you do during an iteration.  The English word experiment derives from the Latin exper which means to try.  Startups use conversations, networking, presentations, and prototypes to try out their product and their business model on potential customers.  There is a lot of accumulated wisdom on different types of experiments and how to run them efficiently.  Experiments create learning experiences for entrepreneurs which lead to better plans.  To experiment is to engage in trial-and-error, or perhaps, more accurately, trial-and-learn.

One of the words repeated in this post is learning.  Entrepreneurship is a process of learning about customers and markets and using that learning to build a profitable business.  The startup methodology outlined here is about how to orchestrate the learning process.

Monday, December 17, 2012

A Startup In Images

What do you do?  Do the build-measure-learn loop, or in the words of Pearl Jam, "Do the evolution, baby!"

from Eric Ries.  Define an idea.  Build something to test your idea.  Measure customer responses.  Modify your idea based on what you learned.  Repeat the process.  The faster you move through the loop, the more your learn, the more likely you're idea will succeed.

What do you measure?  AAARR!, short for pirate metrics for web/mobile startups, which is short for acquisition, activation, retention, revenue, referral.

From Dave McClure.  This is pretty self-sufficient except for activation which is a happy and positive first experience with a product or service.  First impressions are pretty important.

How do you get to startup Nirvana?  Four steps (to the epiphany)

From Steve Blank.  Take a set of hypotheses about your business and turn them into facts by interviewing, presenting to, demoing to, and, in general, getting intimate with customers.  WARNING: startup founders will be first fired in the execution phase because they're not genetically programmed to execute.

How do you represent your hypotheses?  As a business model.

From Alexander Osterwalder.  A de facto standard for representing a business model is in terms of the nine concepts on the above graphic.

Startup Project Management in Four Dimensions

Management of a startup can be thought of as a four dimensional challenge-- the challenge to develop a valid business model, to find and develop customers, to develop artifacts that test the business model, and to perform those tests.  Each of the four dimensions require attention and management.

The goal of a startup is to create/find a good (valid) business model.  The customer is the ultimate judge of the business model through his buying behavior (and buy signals).  Experimentation is the means to discover more more quickly and more early in the process whether the business model is valid.

The Business Model.  A valid business model is the ultimate goal of startup.

Customers.  When developing a startup, you may not need a lot customers at any given point to test your business, but over the course of your project you may "use up" a lot of customers.  Your initial ideas ideas may be flawed and initial prospects may not come back to you after an underwhelming initial experience.   It's important to embrace this reality and be willing to go through a good number customers.  

Customers need to be continually recruited to test out the business and this "recruitment" process needs to be managed.

The "Product".  The product is re-framed as the artifacts needed to test the business model.  The "product" eventually becomes the real product.

Experiments.  Experiments are the means the learn from customers about the busines model as presented by the "product".

Friday, December 7, 2012

Uncertainty and Entrepreneurship

One of the fundamental characteristics of startups which drives thinking about methodology and whether you actually want to do one is uncertainty.

Here's Eric Ries:
A startup is an organization dedicated to creating something under conditions of extreme uncertainty. 

And Steve Blank:
In a startup the search for a business model is chaotic, unpredictable and uncertain. 

No plan survives first contact with customers.

Conditions on the ground will change so rapidly that the original well-thought-out business plan becomes irrelevant.  Blog post

Alex Osterwalder:
Business model innovation remains messy and unpredictable, despite attempts to implement a process. It requires the ability to deal with ambiguity and uncertainty until a good solution emerges. -- Alex Osterwalder,  Blog Post

The world is so full of ambiguity and uncertainty that the design attitude of exploring and prototyping multiple possibilities is most likely to lead to a powerful new business model.  -- Alex Osterwalder,  Blog Post

He wasn't talking about startups but I'm remind of what Nietzche wrote:

I say unto you:  one must still have chaos in oneself to be able to give birth to a dancing star

Friday, November 23, 2012

The 4--No, Wait--6 Noble Truths of Entrepreneurship

The central truths of the Buddhist tradition are found in the The Four Noble Truths which begin with No. 1, loosely translated, "Life is difficult."  There should be something similar in entrepreneurship: a statement of the basic truths that apply in all cases.  I've attempted an itemization which I'm fairly certain will hold out and be uncontested for the rest of the day (assuming no retweets).  Here it is:

1.  Most startups don't succeed with their initial plan.  Customers will defy your expectations again and again.  Therefore the most important thing is to develop and test many plans--A, B,C, etc-- until you get to the plan that will work (its out there somewhere).  This is the fundamental startup truth and a lot follows from it. Its similar to the Buddhist "Life is difficult" and could also be stated, "Comedy is easy, entrepreneurship is harder."  But seriously, you need to get to Plan Z.

2.  Engage customers early and often.  The answers, however difficult to obtain, lie with customers.  They may not know the solution, but they live and breathe the problem and the need.  Develop your understanding and relationship with customers as you're building your product or service--not just in the beginning and not just at the end.  Don't disconnect with customers in the middle.

3.  Your real product is not your product; it's your plan.  Describe your plan simply in terms of  your value proposition, customer, market, and revenue model.  This plan is called a business model and needs to be written down before you write any "business plan".  Embrace the language of the business model and get it down on paper.

4.  Testing is the engine of progress.  Test your entire business model (which you just wrote down) starting with the biggest risks: features, customer acquisition strategies, pricing strategies, etc.  Don't just test your product.  How do you do it?  In small batches, with simple prototypes, with a product that has a few core features, with a revised product with one additional feature.  Test frequently so you don't have enough time to develop a lot of stuff.   You waste less time when the unexpected happens.

5.  Iterate fast and frequent.  Startup development IS a series of iterations with the following steps:  document your model, put something in front of customers, and measure the result qualitatively and quantitatively.  Based on the results refine your model or pivot to a different model.  Repeat the process until you run out of money or get to a working plan.  Think rapid cycle time.  The more iterations you can get in, the more pivots you can make, the more plans you can blow through, the more likely you will get to a good plan.

Paraphrasing liberally from Linus Pauling, the best way to have a good plan is to have a lot of plans.

6.  Prefer funding your startup with customer revenue over investor money.  This is the deeper meaning of bootstrapping.