This blog post has been a collaborative effort with Marianne Hollier, Cindy VanEpps, Roger LeBlanc and Anita Rass Wan

Ever jump into a TV series in the middle of a season? Welcome to Episode 5, Season 1 of DevOps Scenarios for Middleware.  For an overview of this riveting blog series, check out the overview blog here.

You can be a lean startup in an Enterprise. Sounds like an oxymoron, right? Yet Enterprises must innovate to stay alive in today’s transparent, interconnected and fast-changing world. In The Lean Startup, Eric Ries defends the position that “Intrapreneurs” [people who are innovating within an established company] have much more in common with the rest of the community of entrepreneurs than most people believe”, that “Entrepreneurship . . . requires a new kind of management specifically geared to its context of extreme uncertainty.”

Let’s chat about how a team working within a large Enterprise can, in fact, be nimble and creative in getting innovations to market.

This scenario starts with a great idea, of course!  Don’t we all have those every day?  The key is to use learning milestones, that is, small steps that have targeted outcomes that can be measured.  So, for example, we believe that the addition of offering targeted loans in our online banking application will cause a 10% increase in loan applications. Our first learning milestone focuses on whether only offering a back-to-school loan to people with school-aged dependents will give a better application rate than offering the loan to every customer.

So we set about to inject some analytics into our backend systems to identify whether a user has school-aged dependents. If the customer does have school aged dependents, we present that customer with a back-to-school loan offer. This is the Minimal Viable Product for our first learning milestone.

We have a hypothesis that if we offer a back-to-school loan to everyone, regardless of our filtering, then the conversion rate will be lower than if we use a targeted offer.

Then we deploy two versions of the updates to a small subset of customers: one that only provides the loan offer to customers who meet the criteria of having school aged dependents and one that provides the offer to every customer.

Our ability to build-learn-pivot quickly is reliant on a flexible as well as quickly deployable infrastructure. Great ideas still require some middleware capability and horsepower to run them for testing with users.

We use A|B testing techniques to present the different versions to different customers of our web application. Then we use a solution that captures customer interactions to help us validate or refute our hypothesis.

From this analysis we get some interesting results. It turns out that when the back-to-school loan is offered to all customers, we get appreciably the same number of loan applications on back-to-school loans, but in the general populace, just making the offer sparks interest in other types of loans, like home improvement loans and car loans, and the number of those loan applications increases.

In parallel, we test what kind of performance impact both the analytics engine as well as the sudden interest in back-to-school loans has on our infrastructure. We are not only learning what our customers prefer, but also whether our infrastructure can support our innovation.

Next we want to test whether we see the same effect when we offer a home improvement loan only to homeowners.  So we scoot on back to the development shop to craft up the next test. We pivot- rinse-repeat until we have a commercially viable solution that has proof that it will meet our business needs goals.

This is when the proverbial hogs go wild. Now we can do everything required to take what we learned and package it up with all the requisite dotted i’s and crossed t’s, integrate it into the myriad of systems in our enterprise, and spit-shine it into production readiness.  And we can do this with confidence that we have iterated the risk out of our solution through use of a lean startup approach.

Tags:  middleware rational software-development-life... devops scenario