Failing for a Scaling

It never ceases to maze me that in this modern era of software and infrastructure, companies that have Saas-based products still fail to plan for success!

I have been using a home-budgeting product for a few years now. They used to have a desktop version and iOS app all tied together using Dropbox. It worked fine, but was clunkier than they would like - and they could not control some of the moving parts. I also suspect creating desktop versions was becoming more and more expensive compared to a portable web-app.

So they did lots of work and built their new “cloud delivered” SaaS version. To their credit they left the existing version available in “classic” mode, and you could upgrade to the new SaaS version at your leisure.

So they launched - just after Christmas time. I can imagine the discussions they had in advance…

It will be the holidays, not many people will convert.

and

We don’t expect a high initial uptake as people will take time to switch over.

And so forth…

Cue launch day - site immediately swamped with signups and fails (because everyone is at home and has nothing better to do than update some software). To their credit they quickly implement a token system to register and then get a link to signup when capacity is available.

Link comes through a few hours later, so I signup. Part of this process is, naturally, the ability to migrate all your data to the new system. And of course THAT part of the system was totally down - for HOURS.

Now to their credit, they had an excellent status dashboard, and communicated pretty well. But this does not get away from the fact they totally botched their capacity planning and contingency plan. This is made even worse by the fact they are using cloud-based platforms (e.g. Heroku) for their software, so should have been able to spin up LOTS of capacity IN ADVANCE and a minor marginal cost.

It is 2016 and I see the same mistakes being made:

  1. Not enough initial capacity.
  2. Poor strategy for expansion.
  3. ‘Surprise” at the demand for consumer products.

There is simply no excuse for this - because the consumer experience is king and it should be easy, quick and reliable for me to sign up to your service. After all, if you want my money (i.e. your revenue!) you need me to be able to sign on!

So here is my really simple scaling rule to live by:

You always need more capacity than you think - so go 4x.

Whilst many people apply a 1.5x or 2x factor to “educated guesses”, I am FAR more liberal when using an IaaS with a PAYG model - just go 4x and sleep well. You can easily scale down a day or two later and the marginal cost is negligible compared to customer experience and market reputation.

Welcome to 2016 - no excuses for “failing scaling”.