You’re On the Cloud, But Are You Really Leveraging the Cloud? Lessons from Digital Native Businesses

How does the cloud help speed innovation? What’s the difference between being on the cloud and leveraging the cloud? Is it possible to lower development costs and decrease time to market?

To answer these questions, it’s helpful to turn to digital native businesses (DNBs) as a reference. These B2C companies were “born in the cloud,” and many have more than a decade of experience extracting value from cloud services.

DNBs succeed or fail by the quality and speed of their innovation—making digital apps that their customers truly value, while beating competitors to market. Their product and software development teams strive to maximize the use of their budgets and engineering time to create truly differentiating features that matter to their users.

Because many DNBs began as technology startups, they needed to hone their core value propositions to secure venture capital funding. This process forced them to define their differentiated value to the consumer. That might be the convenience of having food delivered through a network of restaurants and drivers, the price and availability of secondhand goods, or a delightful way to consume content.

DNBs also need to continuously improve their productivity and engineering velocity. One way they do this is by shifting commodity capabilities to the cloud. In contrast to differentiating features, commodity capabilities are those the DNB’s customers take for granted. Examples of important but undifferentiated capabilities include a system’s ability to register new users, issue an alert when an order is ready, and scale the backend database. These features usually aren’t part of a customer’s core buying decision, so by leveraging these commodity capabilities, DNB’s can focus their development teams on innovations that truly differentiate and keep customers returning for more. And, because these functionalities can be quickly implemented and optimized, development teams can immediately reduce time spent in ongoing maintenance and improvement costs.

In 2009, when AWS introduced the ability to automatically scale computing power up and down based on demand, this was seen as revolutionary. Companies that adopted this service saved budget, time, and precious engineering effort from their development cycles. Over time, many additional cloud services were added and moved “up the stack” with even more advanced capabilities.

These services included the automation of security, governance, and compliance; support processes for developing and testing software; artificial intelligence (AI) and machine learning (ML) platforms; and tools to help with a myriad of value-adding capabilities, including AR/VR and robotics. For example, Duolingo, an AI-powered language-learning platform, uses the PyTorch framework on AWS to present custom algorithms that provide customers with tailored lessons for 32 languages ranging from Arabic to Welsh. These custom models use anywhere from 100,000 to 30 million data points to make 300 million daily predictions, such as how likely it is that a user can recall a particular word and answer a question correctly.

Duolingo’s system makes these predictions using deep learning, a subset of AI and ML, that analyzes the number of times a user has seen a word, how many times they’ve gotten it correct, the modes under which they answered correctly, and how long it’s been since they’ve practiced it. Using these predictions, the app then injects the word into the curriculum in a context the user needs to master it.

When Duolingo started in 2009, it used traditional cognitive science algorithms to teach foreign languages as part of a translation project at Carnegie Mellon University. But those algorithms couldn’t process real-time data to create personalized experiences that keep users interested in the content.

Deep learning tools have also helped Duolingo improve the accuracy of its predictions to deepen customer engagement. After implementing these tools, Duolingo found that the number of users who used the service, then returned the second day increased 12 percent. Duolingo now has 300 million subscribers, and it continues to depend on the AWS cloud to increase the platform’s speed and scalability and expand the types of predictions it can make.

As Duolingo’s experience demonstrates, the cloud now offers an increasing range of capabilities. These capabilities offer three main advantages:

A product team’s primary focus is to create differentiated product value that will find a big market and make their company a success. Although infrastructure is critical, moving the procurement, design, implementation, and maintenance of hardware and software to the cloud, can help organizations reduce time to market. Most organizations understand that the cloud infrastructure includes physical aspects such as a data center, machines, and storage. However, the most successful DNBs also recognize that software infrastructure represents an even more critical leverage point for increased velocity and quality.

The AWS cloud now provides a continuous integration and delivery (CI/CD) pipeline that allows every developer in an organization to use the same process to test their code for defects. This also ensures that code written by different teams is integrated and works well together and that it’s all staged to deploy at the same time. Once deployed into production, the workload will be monitored, and it will automatically scale up or down to meet demand. In years past, the budget and headcount resources required to create and support these capabilities represented an industry average of 15 percent of the total engineering and operations budget. These functions sometimes were friction points that delayed the deployment of new revenue-generating features. In our experience some best-in-class DNBs now devote very little of their engineering and operations resources to these supporting capabilities, while dramatically decreasing their time-to-market and improving their customers’ experience.

That’s been the case at Coinbase, a digital currency wallet and platform provider with 30 million customers. The San Francisco-based company used AWS Step Functions to automate and govern the deployment of new software features and updates while also better protecting users from cyberattacks. Coinbase not only deploys successfully 97% of the time, but it has also cut the time needed to add new accounts from days to just seconds and significantly reduced the number of its customer support tickets.

Similarly, Freshworks, a San Mateo, Calif.-based company that develops SaaS solutions for small and midsized companies, has supercharged its marketplace with AWS Lambda, a platform that helps developers build and run plugins while managing cost by scaling up or down based on the demand for only that plugin’s function. In just one example, Freshworks agents now resolve customer support tickets in half the time as before.

Insight from data that can help better serve customers is critical for any business. Perhaps no one understands this better than DNBs, which serve the most demanding customers of all: consumers. That’s why these companies place such a high premium on building features that add customer value. While DNBs are deeply technical, the best ones are also ruthless about focusing their time on creating differentiating rather than commodity capabilities. They understand that their consumers care most about real-world benefits, such as gaining reliable movie recommendations, help finding a sports bar in a new city, and insightful restaurant reviews.

Highly-relevant and personalized experiences are driven by the use of AI/ML tools catering to multiple levels of expertise. 1) At a foundational level, specialized developers have the flexibility to customize the underlying software framework itself. This is analogous to building one’s own car engine. 2) The vast majority of companies will build, train, and deploy their models on top of a framework. Like choosing a pre-built engine and then tuning it for your needs. 3) To maximize speed however, companies can deploy a prebuilt model for specific use cases such as fraud-detection or personalization – analogous to buying the ‘sports edition’ or ‘cold weather package’ when you already know your requirements.

Personalization with AI/ML can be especially powerful. Intuit, the financial software company, used the Amazon Personalize service to quickly design and launch a recommendation engine for its Mint consumer budget tracking and planning app. Similarly, Keen, a maker of boots and other outdoor footwear, uses the same Amazon service to track customers’ browsing and purchasing histories to make shopping recommendations. Keen’s test emails using the recommendation feature have delivered a revenue increase of nearly 13 percent. And Ably, a South Korean startup in apparel ecommerce, uses AI to make personalized recommendations on its app’s front page. Ably says the recommendation engine, which taps individual customers’ browsing and purchasing histories, has empowered the company to build sophisticated AI capabilities without any prior experience with ML technology.

Combined with operational excellence, these new levers and capabilities enable faster innovation. To borrow from Isaac Newton:

Mass represents the total resources (budget and people) devoted to developing the differentiated revenue-generating product capability. Acceleration represents the development process and operational environment. Greater force results from maximizing the resources available for differentiated activities, as well as from increasing the velocity of engineering.

Every digital business can benefit from the increased pace of innovation the AWS cloud enables. Here are four recommended starting points:

A visualization of these three factors is shown below. The horizontal X-axis indicates the level of effort, expressed as a dollar cost. The vertical Y-axis represents a subjective assessment of the project’s risk. The circle sizes show the budgets that might be diverted to innovation.

This example uses an assessment of the current state of the organization. This same methodology should be used on a recurring basis. After all, in retrospect, it’s evident that an early adopter of autoscaling in 2009 had a competitive advantage over others that clung to manual provisioning.

A forward-looking perspective can be applied to new projects, as the product and engineering teams discuss the roadmap, levels of effort, and goals. The teams should identify the truly differentiating capabilities and look for opportunities to minimize commodity work. This straightforward exercise can dramatically accelerate a project.

Digital native businesses create innovations that change our everyday lives. They revolutionize the way we shop, entertain ourselves, get around town, and much more. In the process, DNBs have learned that simply running innovations on the cloud isn’t enough. They can teach us valuable lessons on how to leverage the cloud to innovate with great speed.