Improving analytics delivery speed with Little's Law

I am currently reading Lean Six Sigma for Service to find ways to apply Lean and Six Sigma concepts to my profession as an analytics consultant, both for my company and our customers.

Early in the book, Little's Law jumped out as a simple but useful structure for organizing my thoughts around increasing velocity and agility of response to a business' analytic needs. Everybody wants that, right?
 

Lean Six Sigma for Service defines Little's Law as:

• Lead Time is how long it takes you to deliver your service or product once the order is triggered
• Work-in-Process (WIP) is how much work is currently waiting to be completed
• Average Completion Rate is how much work can be completed during a unit of time (day, week, etc. -- this would correspond to the units for Lead Time)

In the next sections I will brainstorm ideas that may help reduce the Lead Time for analytic solution delivery using this formula.

Ideas to reduce Work-in-Process (WIP)

People

• Solution-specific training that preempts questions and increases awareness/adoption of solutions that already exist
• General user training for analytics platform of choice (both this and the previous example can be mitigated with very intentional design)
• Designing solutions that match the capabilities and comfort level of your audience

Process

• Increase quality of problem definition and solution design; learning the business well enough to anticipate future needs when building related solutions
• Granting rights and supplying training to business people to do their own solution creation, beyond basic self-service
  

Technology

• Platforms that enable users to answer their own unique, unanticipated questions (beyond just filtering) and create solutions using governed data or personal data

 
Also in the Process category, the book explains creating a Pull System to minimize WIP, by which someone controls the release of materials into the WIP queue. A maximum capacity is designated for WIP, and nothing new is released into WIP until the current WIP queue is under that cap. Think of this as the old: "I'd be happy to do that for you, but what should I stop working on to make time for it?"
 
One benefit of the Pull system is that it is an opportunity to prioritize or triage based on whatever factors you want to optimize, such as expected business value. When a work item is completed, the next item to be released could be the one with the highest "score", per your rating system. The mechanism would be wasted if you settled for a simple First In First Out (FIFO) approach. (It reminds me of the concept of the backlog, from Agile.)

Ideas to increase Average Completion Rate

People

• Developer training on platforms and languages in use
  
• Getting outside expert help for especially tough problems
• Increasing staff size through hiring or outsourcing
• Outsourcing low-complexity work

Process

• Analyze current delivery processes to eliminate or reduce non-value-add steps, e.g., Just Enough Documentation
• Come up with alternate "fast-track" delivery paths where less rigor can be used for solutions with certain characteristics, e.g., small audiences, short shelf life/question just needs to be answered once, minuscule data volumes, no anticipated reuse
• Templates, coding frameworks, whatever can be done to automate repetitive work
• Minimizing technical debt
• Newer data management techniques (and supporting technologies), e.g., Data Vault, ELT

Technology

• Selecting a platform based on the skillsets you already have or the classes of problem you anticipate solving
• Invest in complementary technologies, e.g., data virtualization, data catalog
• Scalable server architecture that will not become a constraint, faster hardware
  

Conclusion

The process of creating analytic solutions is not a black box. Like other processes, there are discrete, predictable steps with bottlenecks and opportunities for improvement.

If analytic delivery speed is something you wish to improve, I recommend first documenting the current state and current processes at your organization. Then consider which investments would yield the best return for your situation, be it training or technology. What, specifically, do they address? (You should ask this regarding any analytics investment.) I surely missed some important ideas above, but I hope they give you some material to get an effective conversation going.