Predictive transactions are the next big technological revolution

In recent years, data has been the hottest commodity in the world. The money has turned to the companies collecting it, the companies analyzing it, and the data infrastructure companies providing the digital plumbing that makes it all possible.

Over the past five years, data infrastructure startups alone have raised more than $ 8 billion in venture capital, with an aggregate value of $ 35 billion.

We know the names of the biggest companies in space; they include Databricks, Snowflake, Confluent, MongoDB, Segment, Looker, and Oracle.

But what are they really for?

Most investors will explain how data can, in theory, be used to derive trends. Others can talk about how data is going to change the world, without filling in the blanks on how.

I don’t disagree. I have worked and invested in data companies my entire career.

But I think they are missing something big. There is a powerful disturbance to come; perhaps the most powerful since computerized transaction processing was invented in 1964. Predictive transaction processing is about to revolutionize the model of the past 57 years of computing and change the way we live, work, and work. to shop and have fun.

For businesses to remain relevant and competitive, they not only need to be able to predict customer behavior and preferences, they also need to rely on predictive transactions to automate most of their business interactions, i.e. – say performing automated actions while selling or serving the customer. .

A new transformative model

Since the dawn of IT, transaction processing has been done in much the same way. The user makes a request, the request is processed, and if you are lucky, then the user’s choices are analyzed.

This is what is happening today on many platforms.

When I buy a product from Amazon, machine learning can be used to make recommendations. But the decision to buy is basically something that I, the customer, have to make. When I browse Netflix it algorithmically suggests content that I would like to watch, but again I have to make the choice to press play.

We call it “artificial intelligence” but I think it’s not smart enough. The real transformation will happen when we move to a predictive computer model.

Picture this: you’ve just come home from work and an Amazon delivery truck arrives at your doorstep, hauling all 25 household items, from dry groceries to cleaning supplies, you’ll need this week, informed by your detailed customer profile. Anything you don’t need (an unlikely opportunity given the improved machine learning) can easily be returned – information that adds to the database that continually improves engine learning and performance. ability to predict your behavior.

The use case is clear: When transactions pass improvement decisions (i.e. in terms of logistics, last mile delivery technology will ensure people get what they want when they need it, alleviating the burden. traffic jams caused by delivery trucks currently hampered by uncertain deadlines and unavailable customers.

Considering Amazon’s sophisticated logistics and data resources, this scenario is not hard to imagine. Amazon has data on your purchasing habits from a lifetime of purchases. It contains your credit card details. And it has the unmatched ability to ship goods quickly on a large scale.

The same can be true for Netflix and other entertainment platforms like Spotify. They know our habits, so why wait until we tell them what they already know before having fun?

As Benedict Evans says, a computer should never ask a question to which it knows the answer.

This, however, is only the beginning. The predictive transaction processing model is not just an opportunity to improve our lives, our existing systems and our business models. It will be essential to unlock the transformative technologies of the future.

Take autonomous vehicles, for example. We will not achieve “level 5” range if the car only has its own built-in sensors to rely on. We need all cars, from human-driven cars to cloud-based learning vehicles, for the risks on the road to be calculated using the data collected by each autonomous vehicle. And we need this calculation to be predictive, to steer our vehicles in anticipation of the dangers that lie ahead. By acting using the predictive, data-driven model, auto crashes can be a thing of the past.

Predictive transactions will become crucial for industries, from DTC commerce and entertainment to transportation, logistics and even healthcare, as everyone stands to reap the benefits of this incredibly incisive insight into their customers and habits.

Put the building blocks in place

There are already companies that are taking interim measures towards the predictive future.

Most notably, there’s TikTok from ByteDance. With $ 34 billion in revenue in 2020, it is the most profitable predictive transaction processing app ever. Open the app and you will be presented with an endless stream of short autoplay videos. As you watch, the algorithm will learn what you like based not on your stated preference, but your revealed preference.

In other words, if you spend more time watching animal videos than people singing or performing stunts, the app will show you more animals, without you ever having to press play or type words in a search field.

Companies that are building today must follow the example of ByteDance and invest and build the key technologies that will move us towards the predictive transaction processing model.

As part of the shift from user-instrumented interactions to decisions made by learning systems and data, we will need to reorganize and redesign the entire technology stack.

For example, we will need improved machine learning models that are more accurate in their predictions, as marginal gains will make a difference when passed through a supply chain. We will also need learning systems that are able to look back and correct previous mistakes, so that mistakes are not made worse.

We’ll also need to replace long-standing sacred cows, such as the J2EE standards that have unpin e-commerce for a generation. Applications based on learning from data are very different from those based on traditional relational database. We will also need new development and debugging tools, such as new low-level programming languages ​​to allow us to query data more efficiently.

Application integration will also increase in complexity as applications will be entirely data driven rather than design driven.

And ultimately, it will require a sea change in the reliability of real-time transaction processing applications. If predictive data is to be critical, we need platforms and products that reduce downtime, enable instant recovery, and have automatic failover capabilities.

The real opportunity

The revolution in predictive transaction processing is imminent. It is perhaps the most exciting innovation enterprise IT has ever seen. When the technological building blocks are put in place and the applications finally arrive on the market, the impact will be immediately felt.

The number of transactions on predictive platforms will skyrocket. There will be huge opportunities to improve the efficiency of existing systems and a lucrative role for the ecosystem of companies that create the middleware that makes it possible. And the enterprise SaaS platforms that dominate today risk becoming obsolete.

So now is the time to embrace predictive transaction processing, and savvy investors will learn a lesson from this new paradigm: it’s time to look ahead and make decisions now about where to put your money knowing what. coming up.

Alfred Chuang is General Partner at Race Capital (Databricks, FTX, Solana, Opaque), where he invests heavily in data infrastructure. Previously, he was co-founder and former CEO of BEA Systems and led its acquisition by Oracle for $ 8.6 billion.