Monday, 4 October 2010

MVP in the Enterprise

Minimum Viable Product (MVP) is a well understood concept in startups.  It is all about distilling your features down to the barest of essentials – those things which really make your product your product - in order to get something out rapidly and be able to quickly iterate guided by feedback.  From working at both ends of the startup-to-mature-business spectrum, it occurs to me that the enterprise could learn a thing or two from its lighter-weight cousins.

To broadly generalise the circumstances I’ve seen:

We’re pretty good at this MVP thing in early stage startups.  Mostly because, as we plan an iteration, we are fully aware that we might not be around for another iteration so whatever goes into this one has to count (in fact it might count for everything).  We think lean.  We’re typically good at serving our customer’s most immediate and pressing needs.

In the enterprise we have the luxury of working with the confidence that, even if the immediate next set of features don’t quite hit the sweet spot, we’ll be around for many more iterations yet before times start to get tough.  We think deep and wide.  We’re typically good at serving a big strategic master plan.

Being able to take a longer term view of a roadmap is definitely an advantage and having to spend time worrying about things like scalability is a nice problem to have, but this doesn’t mean that we can’t also think lean.  In fact some of the most value I’ve added has been in the amount of work I haven’t done...

It can work in big business.

Recently I looked at a system to capture and analyse certain customer activity in order to get an earlier prediction of lifetime value and make data-driven cross selling decisions (which ought to push up average revenue a notch) and, on the surface, that seems like it’s worth the pretty hefty licensing, integration project, and infrastructure costs.  We started off modelling it and running the numbers; at first synthetically and later with real data.  What we learned through this early prototyping is that those features which were initially so attractive made very little difference to our business, however our little robot was uncannily good at identifying fraudulent activity.  So we switched the direction of the project, chose an alternative platform, saved a bunch of money and got some functionality which genuinely benefitted us.  Sometimes you know what the real benefits are going to be up front, sometimes you surprise yourself along the way.

We’re also planning to introduce a pay-as-you-go commercial model for our feed products.  When you add up a metering system to watch client consumption, a dynamic pricing system to set tariffs, a billing engine to produce the required invoices, and some reporting tools to keep an eye on its performance it starts to become a fairly significant undertaking.  Will customers like it?  Will it work like we hope (add top-up revenues to committed subscriptions) or like we fear (cannibalise commitment for shorter term hits)?  The best way forward in these circumstances is to ask what is the absolute minimum amount of work I can do to see if this works?  As it turns out we could do a some basic instrumentation (to get simplified metering), use some static pricing, and do the billing manually.  Net result?  A much smaller piece of work – useable in production with real customers – demonstrating how successful it’s going to be.  Now we can build on this with further iterations until we have the fully-featured and refined version we first dreamed up.

There are many more examples, some of which have worked out just fine – so we’ve kept building on them until they were feature complete, fully automated, and enterprise quality.  Others have failed – so we’ve cut our losses (fail fast fail cheap; writing off a few thousand not a few million) and moved onto something different which did turn out to be a winner.

Just because we have the budget and the appetite to do the whole shooting match right away doesn’t always mean we ought to.

The obvious risk here is becoming too short sighted with your plans.  The secret to making this work is to imagine big but plan small – keep a long term roadmap and have a clear vision of where you want all your products to get to, but take small, tightly scoped steps along that path.  Stop and evaluate frequently, adjust big picture where necessary, rinse and repeat.

2 comments:

Jan said...

Eachen, thank you for the very interesting post. I forwarded it to several colleagues.

Just out of interest: could you let me know which system you were evaluating to predict customer lifetime value? Either her or via e-mail at jan@hitmeister.de

Thanks!

Eachan said...

Thanks for the comment Jan - glad it helped. I will send you some info separately as I try to avoid being seen to endorse or discredit any particular product/vendor here!