10,000 to one - how smart design boosts customer support performance

When you're managing (or planning) a high-growth online service, one of the key considerations is customer support.  While you can increase the capacity of your servers and boost the bandwidth to meet demand, your reputation could fall apart if you're not able to support growing numbers of visitors.

While banks and retailers often plan on a visitor to customer support ratio of 500:1 and a typical ISP may achieve 1500:1, many online services aim much higher.  Here are two essential features of customer support:

• it can help delight customers and win loyalty - especially among those who initially have a problem
• but it contributes the highest share of variable cost to many online businesses

So, customer support is important, but expensive.  And it can be very hard to scale up.  If you want to scale fast you may have to accept that your service level will drop, and that your bond with some of your visitors may begin to fray.  Some organisations with millions of users respond by making it difficult to reach their customer support (when did you last talk to someone at Skype or Google?), while others swallow the costs of 24 hour call-centres and watch margins shrivel.

But with careful management it is possible to give your visitors accessible customer support without impairing growth.  The key is to provide the support, but minimise the chances that each visitor will want to use it!

This is a risky process - get it wrong and your customer support team will be under pressure - but the secret is to use poka-yoke design principles to ensure that visitors don't need any additional help.

Poka-yoke (roughly translated as "mistake avoidance" or "idiot-proofing") involves designing processes that reduce the chances of problems arising during a process - in this case the experience a visitor has of the service they are visiting.  Here are some design considerations that can help reduce the chances that visitors will need human customer support:

1. keep site layouts simple
2. minimise the number of options and decisions in the process
3. use logical steps to break down the process into intuitive components
4. use automated verification and form checking wherever possible and give real-time feedback

With careful design it is possible to manage and predict the load on your customer support resources by applying these principles in a continuous process.  As certain visitor patterns emerge, design can be modified to anticipate and address common concerns or complaints.  And it's worth remembering that visitors prefer not to have to contact support since it costs them time and effort too!

To serve the needs of 10,000 visitors with a single customer support team member, each visitor can occupy only 2.9 seconds of an 8 hour working day assuming that their support requests are evenly spread!  Since they never are evenly spread, and since the typical support request takes longer than 2.9 seconds to deal with, it's worth applying a queueing model to assist with analysis and preparation.

So, even assuming even distribution, if your customer support requests typically take 5 minutes (300 seconds) to deal with, you need to ensure that fewer than 1% of your visitors need help.

When we first launched Bmycharity we had a tiny fraction of the visitors that the service now enjoys, but we were very cautious about offering any real-time phone or online support, relying on webforms and email instead.  But over the years we systematically redesigned the site.  We made real-time phone support prominently available, and we addressed the design of every process element that feedback indicated caused visitors trouble.  We never perfected the process, but we were able to sustain dramatic growth in visitor numbers without increasing our customer support resources - and we regularly received recommendations for our prompt response to support requests.