Tag Archives: google maps

Lasting Impact of Plaid’s Innovation

Six months ago, Visa acquired Plaid for a cool $5.5 billion, instantly making the fintech company a legend among technology startups – and its founders, investors and early employees very rich.

While the money is fun to consider, it’s not my key takeaway about Plaid, whose software provides the plumbing that lets startups connect to users’ bank accounts and has been employed by peer-to-peer payment app Venmo, mobile investing app Robinhood and many others.

As the CTO of a startup – one developing a technology platform for insurance carriers – I’m finding the real topic of conversation among my peers, as well as among my company’s investors, partners and prospects, is Plaid’s technology approach and its ramifications.

Plaid’s decision to focus on application programming interface (API) development vs. application development is a natural starting point for such a discussion. (The main point being that Plaid didn’t set out to build all the financial apps itself; instead, it provided a key interface that others, like Venmo and Robinhood, could exploit.) But there’s more to the story. And while it has already become fashionable to describe certain companies as “the Plaid of (fill in the industry),” I don’t think Plaid will be remembered as the face that launched a thousand API ships.

Not to say that the “API-ification” of the enterprise isn’t upon us. We’re seeing it in insurance and more broadly across financial services, as more processes within and across companies plug into each other via these software interfaces. But that trend started before, and is bigger than, Plaid. Conversely, Plaid’s significance extends beyond its APIs.

So what then can Plaid teach us? What can startups, and the technologists helping build them, learn from this early 2020 success story and carry forward into the young decade?

Here are four key takeaways:

1. Empower Builders

Any time a company develops a technology that makes it possible for others to do something they couldn’t do before, that company has the makings of a hit. Many companies succumb to the temptation of trying to own too much of that innovation’s value rather than putting some of that value creation into the hands of others.

Consider Google Maps. Prior to Google Maps, it wasn’t easy for a company to build a dynamic map into its customer experiences. Yet with the Google release, retailers could put all their locations into their web experience so consumers could find them without having to go to a specialized third party service. Now dynamic maps are an integral part of an array of experiences from retail and restaurants to real estate and travel.

If Google Maps had insisted on being the destination of all things maps, i.e., “Find My Retailer,” “Find My ATM,” “Find My Restaurant,” etc., and owning the entire value proposition, the proliferation of map-enhanced experiences across the internet would not have been as quick or as pervasive.

Even a company of Google’s size recognized that that approach would have put the burden of application-layer innovation on one company, one set of developers and one team of product managers. Instead, Google famously developed dynamic maps to be an embeddable component that can fit into any other application, enabling a variety of developers to innovate for their particular markets and end-users.

That’s something Plaid got right. In developing its APIs, Plaid unlocked banking data that had never been available and usable before, but the company was smart enough to keep the focus there and let others – Venmo, for instance – innovate at the application layer. Plaid’s approach brings so many more companies into the innovation mix, which in turn spurs more A/B testing, ultimately yielding more robust and varied applications – the true benefit of best of breed.

2. Love Your Hacks

Plaid did a lot of things that business school won’t teach you. One of them was embracing the hack. Experience has shown that many of the most successful tech companies have used hacks to get their businesses off the ground and deliver their first positive results. Airbnb and Uber come immediately to mind, and so does Plaid.

This is important, because banks were never in the business of exposing their data in a clean way – they didn’t have nice open APIs with clean documentation. That meant that Plaid had to crack the code on its own, figuring it out for itself every step of the way. How do we get access to Bank of America data? to Chase? to the next big bank?….

The moral of the story is they did it. Sure, initially they did it with workarounds and solutions that were laughed at on forums and dismissed as insecure. But they stuck with it and built big enough engineering and data teams to make the company, its approach and its solutions sustainable over time.

But first came the hack. Before becoming sustainable, before scalability even enters the equation, Plaid was getting its hands dirty showing the value of its work, no matter how unsustainable the approach. In this regard, I see Plaid as embodying Paul Graham’s famous admonition to startups: “Do things that don’t scale.” I would just add: “until they do.”

3. Trust the Size of Your Market (and the Defensibility of Your Solution)

Technologists and entrepreneurs have to develop thick skins. We hear “no” more than “yes,” and often find ourselves answering the same questions: What is the size of your market? How are you going to monetize your innovation? How defensible is your solution?

Here, I have to tip my hat to Plaid. As a trailblazer, the company had to argue for a big market that didn’t exist yet, because no one else was monetizing access to financial data – and the generation of apps that would use that data so successfully had yet to be created.

Of course, Plaid was right. Other companies would use its APIs and multiply their value many times over. But before Plaid was right, it believed. The size of its conviction ultimately enabled it to create and fuel a multibillion-dollar market. And while it believed, it minded its knitting, focusing intently on innovating and letting great software speak for itself.

Did that answer investors’ initial questions about defensibility? At first, probably not, but as the number of successful hacks mounted, and as it became clear that the problem it was trying to solve was sufficiently complex and the competitive landscape it inhabited sparsely populated,  the company earned enough breathing room to deliver each successive, successful result. By 2018, Visa and Mastercard were in on the company’s $250 million raise, and the rest is history.

See also: Insurance Innovation — Alive and Kicking

Getting there took some swagger, perhaps even a little arrogance, that Plaid could solve something no one else had dared attempt. That attitude may have been its best line of defense.

4. Guillotine Your Platform!

As I mentioned, the temptation to try to do too much, to own all the value and innovation at every layer of a solution, can be fatal, and is something Plaid brilliantly avoided. Plaid will be remembered for focusing on APIs and powerful administrative functionality, leaving the user interface (UI) and user experience (UX) layers for others to perfect and deploy.

In this case, Plaid serves as a powerful example for the many “platform” developers across the startup landscape, mine included. Platform developers want to solve it all, but Plaid is helping us not to. They deliberately chose not to provide the full vertical experience of their service, leaving it up to developers outside their company to figure things out for themselves and provide their customers their own distinct experiences.

This “headless” platform model is quickly gaining traction among startups and other solution providers, as well as among big companies hoping to accelerate or complete their digital transformations. These companies don’t want their tech providers to own any portion of the customer’s journey and experience; they just want the value, and they want it expressed natively within their own digital footprint.

That shift, and tech startups’ ability to deliver on it, may be Plaid’s most lasting legacy.

Why I’m Skeptical on Apple’s Future

Facebook is releasing its virtual reality headset, Oculus. It is big, clunky and expensive, and it will cause nausea and other problems for its users. Within a few months of its release, we will declare our disappointment with virtual reality while Facebook will carefully listen to its users and develop improvements. Version No. 3 of Oculus, which will, most likely, come in 2018 or 2019, will be amazing. It will change the way we interact with each other on social media and take us into new worlds—much like the holodecks we saw in “Star Trek.”

This is how innovation happens now, innovation and elsewhere. You release a basic product and let the market tell you how to make it better. There is no time to get it perfect; your product may become obsolete before it is even released.

Apple has not figured this out yet. It maintains a fortress of secrecy, and its leaders dictate product features. When it releases a new technology, it goes to extremes to ensure elegant design and perfection. Steve Jobs was a true visionary, but he refused to listen to customers—believing he knew what they needed better than they did. He ruled with an iron fist and did not tolerate dissent of any type. At Apple, people in one division did not know what others in the company were developing.

Seven announcements Apple made in the March keynote

Jobs’ tactics worked very well for him, and he created the most valuable company in the world. But without Jobs, given the dramatic technology changes that are happening, Apple may have peaked. It is headed the way of IBM in the ’90s and Microsoft in the late 2000s. Consider that Apple’s last major innovation—the iPhone—was released in June 2007.

See Also: Apple v. FBI: Inevitable Conflicts on Tech

Since then, Apple has been tweaking its componentry, adding faster processors and more advanced sensors and releasing bigger and smaller forms—such as with the iPad and the Apple Watch. Even the announcements Apple made this month were uninspiring: smaller iPhones and iPads. All Apple seems to be doing is playing catch up with Samsung, which offers tablets and phones of many sizes and has better features. Apple has been also been copying products (such as Google Maps) but not doing it very well.

There was a time when technology enthusiasts like me felt compelled to buy every new product Apple released. We applauded every small, new feature and pretended it was revolutionary. We watched Steve Jobs’ product announcements with bated breath. However, now I would not even have bought the iPhone a few months ago unless T-Mobile included a large rebate to switch networks. There is nothing earth-shattering or compelling about Apple’s new phones—or, for that matter, any of the products it has released since 2007.

By now, Apple should have released some of the products we have heard rumors about: TV sets, virtual reality headsets and cars. Apple could also have added the functionality of products, such as Leap Motion and Kinect, with the iPhone functioning as a Minority Report motion detector and projector. Apple should be doing what Facebook is doing: putting out new products and letting the market judge them. And Apple should be doing moonshots like Google, which is toying with self-driving cars; Internet delivery via balloon, drone and microsatellite; and Google Glass. Yes, Apple might have failed with the first version—just as Google did with Glass—but that is simply a learning experience. The third version of Google Glass is also likely to be a killer product.

Instead of innovating, Apple has been launching frivolous lawsuits against competitors like Samsung. My colleague at Stanford Law School, Mark Lemley, estimated Apple had spent more than $1 billion in attorney and expert fees in its battle against Samsung. And this lawsuit netted Apple just $158,400, which, ironically, went to Samsung. Apple could have better spent its money on the acquisitions of companies that would give it a real edge.

Will Apple release some products later this year that will blow us away? I am skeptical. I expect we will only see more hype and more repackaging of tired old technologies.

Join Vivek Wadhwa for the Path To Transformation Symposium by registering here.

AI’s Huge Potential for Underwriting

For decades, the insurance industry has led the world in predictive analysis and risk assessment. And today, with the treasure trove of big data available from historical processes, IoT and social media, insurance companies have the opportunity to take this discipline to a whole new level of accuracy, consistency and customer experience.

The actuarial models that were once driven solely by large databases can now be fueled with tremendous quantities of unstructured data from social media, online research and news, weather and traffic reports, real-time securities feeds and other valuable information sources as well as by “tribal knowledge” such as internal reports, policies and regulations, presentations, emails, memos and evaluations. In fact, it is estimated that 90% of global data has been created in the past two years, and 80% of that data is unstructured.

A large portion of this data now comes from the Internet of Things — computers, smart phones and wearables, GPS-enabled devices, transportation telematics, sensors, energy controls and medical devices. Even with the advancement of big data analytics, the integration of all this structured and unstructured data would appear to be a monumental achievement with traditional database management tools. Even if we could somehow blend this data, would we then need thousands of canned reports, or a highly trained data analytics expert in every operating department to make use of it? The answer to this dilemma may be as close as our smartphones.

Apps that Unleash the Power

As consumers, we are no stranger to the union of the structured and unstructured datasets. A commuter, for example, used to rely on Google Maps to get from his office to his home. But with the advent of apps like Waze, not only can he get directions and arrival times based on mileage and speed data, but can also combine this intelligence with feeds from social media and crowd-sourced opinions on traffic. Significant advances in the power of in-memory processing, machine learning, artificial intelligence and natural language processing have the potential to blend millions of data points from operational systems, tribal knowledge and the Internet of Things — using apps no more complicated than Google Maps.

Using apps that harness the power of artificial intelligence and machine learning can provide far superior predictive analysis simply by typing in a question, such as: What are the chances of a terrorist act in Omaha during the month of December? Where is the most likely place a power blackout will occur in August? How many passenger train accidents will occur in the Northeast corridor over the next six months? What will be the effect on my fixed income portfolio if the Federal Reserve raises short term interest rates by .25 percentage point?

Using a gamified interface, these apps can use game theory such as Monte Carlo simulations simply by moving and overlaying graphical objects on your computer screen or tablet. As an example, you could calculate the likely dollar damages to policyholders caused by an impending hurricane simply by moving symbols for wind, rain and time duration over a map image. Here are some typical applications for AI app technology in insurance:

Catastrophe Risk and Damage Analysis

Incorporate historical weather patterns, news, research reports and social media into calculations of risk from potential catastrophes to price coverage or determine prudent levels of reinsurance.

Targeted Risk Analysis (Single view of customers)

With the wealth of individual information available on people and organizations, it is now possible to apply AI and machine learning principles to provide risk profiles targeted down to an individual. For example, a Facebook profile of a mountain climbing enthusiast would indicate a propensity for risk taking that might warrant a different profile than a golfer. Machine learning agents can now parse through LinkedIn profiles, Facebook posts, tweets and blogs to provide the underwriter with a targeted set of metrics to accurately assess the risk index of an individual.


Each individual assessor has his own predilection to assessing risks. By some estimates, insurance companies could lose hundreds of millions of dollars either through inaccurate risk profiling or through lost customers because of overpricing. AI apps provide the mechanics to capture “tribal knowledge,” thereby providing a uniform assessment metric across the entire underwriting process.

Claims Processing

By unifying unstructured data across historical claims, it is possible to establish ground rules (or quantitative metrics) across fuzzy baselines that were previously not possible. Claims notes from customer service representatives that would previously fall through the cracks are now caught, processed and flagged for better claims expediting and improved customer satisfaction. By incorporating personnel records when a major casualty event occurs, such as a severe storm or flood, you can now dispatch the most experienced claims personnel to areas with the highest-value property.

Fraud Control

Integrate social media into the claims review process. For example, it would be very suspect if someone who just put in a workers’ compensation claim for a severe back injury was bragging about his performance at his weekend rugby match on Facebook.

A Powerful Value Proposition

The value proposition of artificial intelligence apps for better insurance industry underwriting and risk management is too big to ignore. Apps have been transformational in the way we intelligently manage our lives, and App Orchid predicts they will be just as transformational in the way insurance companies manage their operations.

How Google Thinks About Insurance

For those of us wondering what Google plans to do in insurance — isn’t that all of us? — it’s worth looking at the company’s Project Sunroof. The project uses exceptionally sophisticated mapping data to determine which homeowners would most benefit from solar panels and, in the process, may provide some insight into how Google is approaching insurance.

To me, there are two key aspects of Project Sunroof. The first is that Google is taking a bottom-up approach that could inform a lot of decisions about insurance (while insurers traditionally go top-down). The second is that Google is being unusually smart about combining layers of information — some proprietary, some in the public domain; some new, some long-available — to produce what my frequent co-author Chunka Mui has, with a little help from me, labeled “emergent knowledge.” (“Big data” is the term commonly used, but data isn’t very interesting, while knowledge is. And the size of the database doesn’t matter. What matters is using developments in technology to look in the right places to find the right data to answer the right questions so that revelations emerge.)

Top-down vs. bottom-up

Insurers typically start with pools of risk. They’re getting much more sophisticated about subdividing those pools into ever smaller groups, but the thinking is still along the lines of “drivers without moving violations who travel 11,000 to 12,000 miles a year in generally suburban conditions.” Insurers will keep getting more and more specific and produce more and smaller pools but are still going from the top down.

Now look at Project Sunroof. Google is modeling the world in three dimensions and using that model to generate information house by house based on totally personalized criteria: on the square footage on the roof that would be available for solar panels, on the amount of cloud cover that is expected to obscure the sun above that house, on the effectiveness of the sunlight that will hit the roof (incorporating calculations based on temperature and on how the angle of the sun changes each day) and on any shade that would be cast on those panels from other structures. Although the article doesn’t say so, I assume Google calculates potential savings on solar based on the rates of each local utility. In any case, there are no pools in sight for Google — unless you want it to tell you about those in the backyards.

That same model of the world could be the basis for a house-by-house, car-by-car, person-by-person approach to insurance for Google. And, if this approach works, Google will gain the sort of information advantage that has proved to be almost impossible to overcome. Even the largest insurers would have a hard time spending the money that Google has to map the U.S. by having cars drive every single street to take pictures and collect data, by making a series of acquisitions of data providers and by employing a small army of people to manually fix errors and update maps — and Google would still have a years-long head start in developing its model of the world. Microsoft has thrown billions of dollars at search engines, but even Microsoft couldn’t overcome the fact that Google’s dominant share meant it was always learning and improving faster than Microsoft’s Bing. Apple’s map services were ridiculed, by comparison with Google’s, when Apple launched them in September 2012. Apple is now at least in Google’s ballpark on mapping, but no insurer can come close to Apple’s resources — a $646 billion market valuation and $202.8 billion of cash in the bank. That’s “billion,” with a “b.”

Emergent knowledge

Google obviously begins with a huge asset because of its prescient decision years ago to map the entire U.S. and because of the recent work that has made that map 3D.  But Google is also taking data wherever it can get it.

I know from some work I did at the Department of Energy in 2010 that national maps of sunlight have been available for years, and they have surely become far more detailed as the interest in solar power has spread, so I assume Google didn’t have to generate those maps on its own. Temperature maps are also in the public domain. (Especially high or low temperatures degrade the performance of solar panels.) Those maps will become increasingly granular as they incorporate data from smartphones and other widely used devices that can act as sensors — temperature will no longer be what the weather station reports from the Detroit airport; temperature will be known house by house. Overhead photos from satellites and, in some cases, drones are widely available, so Google can use those to check square footage of roofs, to see which direction the solar panels would point and so on. Google can collect information on rates from state utility commissions, where utilities have to make regular filings.

It’s easy to imagine Google layering similar types of information onto its map of the world for insurance purposes. In response to the federal Data.gov initiative, governments at all levels are making more information available digitally, so Google could incorporate lots of data about where and when accidents occur, where break-ins happen, where and when muggings occur and so on.

Google could incorporate private work that is taking a 3D approach to flood risk (whether your house is three feet higher or lower than the average in a neighborhood can make all the difference) and is being much more discriminating about earthquake risk. Google could add information, from public or private sources, on the age of homes, type of pipes used, appliances, etc. to flesh out its understanding of the risks in homes.

And, of course, Google will have lots of very precise information of its own to add to its model of the world, based on, for instance, what it knows about where your smartphone is and can infer about where you park your car, where and when you drive, etc.

Once you take all this information and map it to such a precise model, there will surely be some non-obvious and highly valuable insights.

WWGD: What Will Google Do?

Looking at Project Sunroof still doesn’t say a lot about how Google will attack insurance. Will it just sell increasingly targeted and valuable ads? Will it sell leads? Will it become a broker? Will it do more?

But I think it’s safe to say that, whatever Google does, its starting point will the most sophisticated model of the world — and that model will always be improving.