Learning as we go

Learning how to scale out a support system in a small business is one of the most challenging things that I have ever done. MXroute support wait times are high, and the workload grows at a rate which exceeds my personal ability to answer the questions. I can look at this as a failure and retreat, or I can see it as an opportunity to learn and improve.

When scaling out a support infrastructure, it's easy to learn the ways not to do it. You experiment, observe the result, and quite often your theory doesn't hold up to reality. As discouraging as failure can be, no one articulated the opportunity it presents better than Thomas Edison:

I have not failed. I've just found 10,000 ways that won't work.

There in lies the opportunity. Each failure is simply a lesson in what not to do. Here's what we've learned doesn't work:

  1. A staff-created Knowledgebase.
  2. Exposing common questions prior to accessing the contact form.
  3. Using Slack to siphon off quick questions and prevent tickets.

So once again we're trying a new strategy. We're going to change the support workflow in pieces. I'll go over the rough plans for that now.

This is to allow you to ask the questions in your own words. Perhaps even duplicate questions because one of the existing ones may not be worded in a way that it looks like it could answer your question. This creates a community-assisted database of material that can be used to expose better answers to questions prior to opening support tickets. That brings us the next step.

  • Second, we're going to launch a chat bot.

This is a heavily trained bot that will be intended to help you identify the right answers to your questions without ever needing to speak to a human.

Now, we value human support just as much as you do. We also recognize that having a human answer every question is simply not a viable strategy at our scale. The more questions we can answer without a human, the more available we can be when you actually do need to talk to us. It's about maximizing the quality of human interactions by automating the repetitive ones.

Rest assured that MXroute support is not dead, we're just struggling to find the right answer. We've found a bunch of ways that don't work, so we're that much closer to the ones that do.