Aditya Ganjam is Conviva‘s Chief Product Officer and Co-founder.
Today’s operational intelligence solutions miss the mark on the most essential thing: quality of experience (QoE).
Rather than focusing on QoE, businesses prioritize system health and performance, defined as quality of service (QoS). The reason? Legacy observability tools can’t collect the necessary information or compute QoE in real time and connect it to QoS.
QoS is important in evaluating system performance and monitoring the four golden signals: latency, traffic, errors and saturation. However, measuring QoS is no longer sufficient to remain competitive. QoE must be the focus of operational intelligence because it’s more holistic and prioritizes end users. It’s measured through every user interaction and resulting system or application response.
QoE must be computed based on patterns of events collected from endpoint devices, which requires the backend analytics stack to perform stateful computation. Unfortunately, legacy observability tools cannot do stateful computation. Below, I’ll discuss the differences between stateful and stateless analytics, why stateful is harder and how the right tools help businesses deliver excellent experiences.
Stateful Versus Stateless
Stateful means the timing and order of events matter, and the data processing system must remember past events—the entity’s state—for every session to process future results effectively.
Let’s look at a few examples.
• Retail: How long before a user abandons a cart after adding an item and initiating a search for a second item? To answer this, we examine the state of the cart (an item was added), when the new search was initiated and when the user left the site.
• Travel: How many users start a search, select a flight and don’t complete the booking? To answer, we follow a similar process: evaluate the state (flight selection), what happened during the booking process and why they didn’t finish.
• Media: How many users open a streaming app, search for content and don’t watch anything? We evaluate the state (they opened the app), what happened while searching and why they didn’t choose a program.
• Finance: How many users abandon a stock trading app after a purchase fails? We remember the state (a stock purchase failure occurred) and see that the user abandoned the app. We can also compute how many users who retry and succeed.
Stateless means the timing and order of events don’t matter. Data or stream processing systems don’t need to remember past events to calculate future results. In the finance example, we could compute the number of stock purchase failures as a count of failed events. However, with stateless computation, we couldn’t compute abandonment or retry success, two far more valuable metrics than a raw count of failures.
Remembering past events is called state, defined as the condition of some entity at a particular point in time. We can talk about the state of an application, microservice, database, network or video player.
For a few more examples, consider the following: Stateless metrics for transactions may include numbers of checkouts or numbers of failures. Stateful metrics instead include time to complete checkout, the percentage of successful retries, the percentage of abandonment and the percentage of abandonment due to slow checkout times.
The Power Of Stateful Metrics
In legacy observability tools, stateful analytics are processed in the agent, so the backend doesn’t need to do it. But QoE is stateful and highly customized, so it must be computed on the backend, which today’s tools cannot do. Businesses struggle to find problems and triage them quickly without proper visibility into the experiences of actual users, who might be having issues while backend systems look fine. Consequently, they discover issues when user complaints occur instead of identifying problems beforehand.
On the other hand, stateful metrics offer context and inform businesses about the experiences that matter most: sign-up, login, content discovery and conversion. The “moments that matter” are checkpoints in a user’s journey that affect revenue, such as
• Hitting “buy now” on a new product.
• Pressing “subscribe” on a streaming service.
• Submitting an online food order.
The problem is that today’s tools can’t compute real-time stateful, QoE metrics. When businesses lack visibility into experience—the bridge between performance and engagement—they don’t know whether the moments that matter are running smoothly or frustrating users.
Businesses can define many QoE metrics. To achieve excellent QoE and keep users happy, companies must monitor these in real time, take real-time actions when they have issues and track them over an extended period for continuous improvement. Thus, the data processing system must be able to compute and diagnose QoE metrics in real time.
While stateful analytics can be done with existing platforms, it’s inefficient and impractical. To solve this, a new abstraction and data processing architecture is necessary.
Start With The Customer
Without proper monitoring tools, companies risk platform outages that anger and frustrate consumers. They’ll complain on social media or in-app store reviews, leaving companies scrambling to resolve the issue. But that’s only part of the equation and, increasingly, not the most crucial part. While system outages happen a few times a year, bad user experiences happen daily and are significantly more impactful than an outage.
Businesses must prioritize the entire user experience, seeing the platform through users’ eyes. It doesn’t require a complete system overhaul, only a shift in technology and methodology. In the words of Apple co-founder Steve Jobs, “You’ve got to start with the customer experience and work back toward the technology—not the other way around.”
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
Source link
