Main content

Tuning application performance to business outcomes

Phil Wainewright Profile picture for user pwainewright January 27, 2015
When mobile and web apps trigger processes across multiple systems, enterprises need the right metrics to tune application performance to business outcomes

businessman hand touching digital tablet © peshkova -
Enterprises today are connecting a growing number of web and mobile applications into an already complex infrastructure of back-end systems. Monitoring the application performance across so many moving parts is a big job — much of it still executed in established technologies such as Java and .NET, but also extending into web and mobile environments such as PHP, Python and node.js, and spanning public cloud as well as private assets.

What really matters about all this technology is whether it delivers the real-world results that people depend on. Raw performance metrics aren't useful unless you also monitor user experience and business outcomes, argues Jyoti Bansal, founder and CEO of application performance monitoring (APM) vendor AppDynamics.

If you had a slowdown in your app, how does it impact your revenue? If you have a new version of your app, is it improving the conversion rate or decreasing the conversion rate? Are you seeing more user engagement from that new version, are you seeing more revenue?

We're calling it application intelligence, where we are evolving into, from performance monitoring.

Business process forensics

Imagine for example a retailer notices a drop in online conversion rates when its website runs slow. It doesn't want loyal customers to go elsewhere because of a bad experience, so it uses monitoring software to detect when regular customers abandon their purchase during a site slowdown. It then automatically sends those customers a five-dollar discount code to bring them back.

Another example is a bank with multiple systems inherited from a dozen different acquisitions over the past decade or so. In some cases, setting up a business account for a new customer can take hours or even days more than it should. Following the transaction across all those different business systems can isolate the source of the delay so it can be fixed. "This customer is calling it business process forensics," says Bansal.

Marketing automation is another field that can benefit from these investigations. Monitoring application performance can shed new light on why digital campaigns go awry, he explains.

Why is the conversion rate slowing down? There could be several reasons. One is that the site is not designed properly or the app is not designed properly. The second reason is it's not performing properly or there is latency in the stack. That is where we come in.

The marketing analytics products, you will see data on when the user is clicking something [but] you only see data at the edge. If you click to provision a new phone line and that goes and triggers calls into your partners or your service providers, how do you correlate what the user is doing with what is happening in the downstream delivery systems?

That's what AppDynamics brings in. You can correlate what's happening from the user click down to every business system that is involved in processing that click that happened.

Putting the 'biz' into BizDevOps

Jyoti Bansal CEO AppDynamics
Jyoti Bansal, AppDynamics

This correlation of technology behavior to business outcomes is starting to affect how IT staff are compensated, says Bansal.

A lot of the time we assume the marketing guys need to understand the revenue impact and the IT guys only care about the technology. That's started changing now.

IT guys, their bonuses will be based on revenue, on conversion rates and user engagement. That's the big shift that's happening. The boundary is getting blurred.

This is giving rise to a new approach to application development and monitoring, says Bansal. The emergence of fast-moving cloud and mobile apps has already forced developers and operations staff to work more closely together — a phenomenon known as DevOps. Now business teams are getting closely involved too.

We are starting to call it BizDevOps. There used to be an assumption that ops only need to care about uptime and errors, but that's not true any more. Ops and dev, if they know about the business impact, they can optimize in a much faster loop, and they can optimize the right things.

That's what our customers are finding really exciting and really interesting. It bridges the three together — the operational data on what is going on, the performance and availability and user experience, to the business data on what is the revenue, what is the conversion rate.

So the three groups are much more aligned, there's faster innovation, there's better business alignment. You can prioritize what is important to the business, what matters and what doesn't. It eventually drives a better experience for the consumer.

Land and expand

AppDynamics is one of a new generation of application performance monitoring vendors that collect detailed metrics into big data stores (based on Hadoop in its case) for analysis. Its rival New Relic is the best known after a successful IPO last month. Whereas New Relic is a SaaS-only vendor that works mainly with next-generation infrastructure, AppDynamics' core focus is on Java and .NET systems. For more background, see my previous interview with Bansal from last year.

With AppDynamics also in the frame for an IPO at some point, it was inevitable that last week we'd talk about the Box IPO and its focus on building market share at the expense of profitability. Bansal said that gaining enterprise customers often meant spending a lot in the beginning to reap returns later on.

A lot of these new businesses, like AppDynamics as well, we follow this model of the 'land and expand' sales strategy.

When we go to large enterprise, we are not trying to sell them for the whole enterprise upfront. We are looking at one business unit, one department, one set of applications. And then we are going to a second one, a third one ...

The first time we acquire a customer, it could cost us much more, but there would be so many more opportunities in that large enterprise if we make the customer successful the first time.

This approach is the antithesis of how traditional IT management vendors have gone to market, he added. The likes of IBM, CA, Compuware, HP and BMC offer products that often take weeks to deploy, whereas AppDynamics can be provisioned via self-service, he said.

We don't need to come and deploy for you. We prefer not to. It's better for us to teach our customers to fish instead of fishing for them all the time.

The traditional business model that you have these large perpetual licenses with a heavy element of professional services to implement, lots of complexity — customers don't want that any more.

Customers want to start small, they want to get success from it. They want self-service, easy deployments where they can learn and do things. They want it to be more organic expansion.

That's the new way how even some of the largest enterprises want to buy and consume software. That's such a massive shift, you have to almost change the DNA of some of these other vendors to fit into that.

My take

Every new wave of computing seems to give birth to a new breed of IT management and monitoring vendors. Mainframes and client-server each have their own. Now web, cloud and mobile are attracting a new generation.

It's all too easy when this happens to end up creating a series of management silos that only look after their own segment of the enterprise IT landscape. In an era when enterprises are already faced with the challenge of joining up data and process flows across a 2-speed IT infrastructure, it's essential in my view to be able to manage performance across the entire landscape. Not doing so is to focus on the technology alone and ignore the wider business context that it was built to serve.

I therefore took the opportunity of this second conversation with Bansal specifically to hear more about use cases that bridge these gaps. He produced some good examples but it struck me that very little of this has been productized — these are all custom implementations of the technology tailored to the needs of individual organizations.

We are not yet at the point where, for example, someone looking at a marketing automation dashboard can routinely include application response times as a metric when evaluating results of A/B campaign testing, or create an alert that automatically pauses outbound campaign messaging when the website is running slow. To achieve that, they first need to persuade their development and operations colleagues to custom-build a way to collect those metrics and feed them into the marketing automation system.

AppDynamics is clearly operating in a fast-growing niche that will quickly become mainstream. At the same time, I suspect it's barely scratching the surface of what's achievable when its capabilities get packaged into APIs and tools that simply plug in to other applications. I'd like to see it do more to foster the growth of that ecosystem.

Image credit: © peshkova -

A grey colored placeholder image