More on this book
Community
Kindle Notes & Highlights
The first method involves comparing the two datasets to figure out which new songs are related to the ones you like. For instance, suppose someone made a playlist with eight songs on it, and seven of those are in your library. You probably like this type of song, so Discover Weekly might recommend the one song that isn’t in your library.[32] This technique is called “collaborative filtering,”
These snippets of code that let you borrow another app’s functionality or data are called APIs, or application programming interfaces. In short, APIs let apps talk to each other. Let’s look at three main kinds of APIs.
The first kind of API, which we’ll call “feature APIs,” lets one app ask another specialized app to solve a particular problem, like calculating driving directions, sending text messages, or translating sentences.
The second type of API, which we’ll call “data APIs,” lets one app ask another app to hand over some interesting information, such as sports scores, recent Tweets, or today’s weather.
The final kind of API, “hardware APIs,” lets developers access features of the device itself.
Despite the potential business risks, using a specialized company’s API is easier, more reliable, and often cheaper than trying to build it yourself.
With this single sign-on (SSO) API, any app can let users create accounts by linking their Facebook profiles.
publishing APIs is a great way for companies to gain data and usage, and using APIs helps apps save development time and offer better features — but it isn’t without risk.
When companies perform A/B tests, experimenters report how one version changed a particular metric compared to the other version. They also report a statistic called a p-value, which shows the probability that the difference they observed was due to chance.[99] Usually, if p < 0.05 (i.e. there’s a less than 5% chance that the difference was just random), they can assume the change was meaningful, or “statistically significant.”[100] Otherwise, they can’t be sure that their results weren’t just dumb luck.
Android powers over 80% of the world’s smartphones.
Google insists that phone manufacturers prominently feature Google Play, Android’s app store, on the home screen of any Android phone.[141] This is designed to drive more users to download apps from the Google Play marketplace. And whenever someone buys an app or makes an in-app purchase, Google takes a 30%
The kernel is a bit of software that lets apps talk to the device’s hardware, like by reading and writing files, connecting to the keyboard and Wi-Fi, and so on.
One big reason that game and app creators love in-app purchases are that they’re pure profit: once you’ve built your app or game, it costs basically nothing to give a user virtual items like outfits or geofilters (in other words, the marginal cost is zero).
There are two big reasons app makers are turning to subscriptions. First, they provide a steady and reliable source of income, whereas relying on one-off purchases leads to big spikes in revenue when you release a new update and big crashes at other times. Second, customers who subscribe tend to stick with the app longer (probably because they feel they have a long-term relationship with the app instead of a one-off interaction), meaning they’ll pay the company more money throughout their lives.
while most people won’t pay anything, the small fraction of people who use a certain app the most are often willing to pay a lot of money. Economists call this the 80-20 rule or the Pareto principle: 20% of your customers will generate 80% of your revenue, and 80% of your customers will generate the other 20%.[262]
In short, the freemium strategy is this: give away the app for free to draw in a massive number of users,[267] find the “power users” who love the app, and charge them — either once or via a recurring subscription — for extra features.
Since these hyper-targeted ads are more relevant to you, you’re more likely to click on them than non-targeted ads.[279] More clicks lead to more purchases, so targeted ads help advertisers make more money.
“if you aren’t paying for the product, you are the product.”
Over half of people, when searching for products to buy online, now start their search on Amazon instead of Google.
fewer and fewer major apps have banner ads nowadays, probably because ads are annoying and take up valuable space.
people only intentionally click on banner ads 0.17% of the time[299] — that’s roughly 1 out of every 600 ads you see.
if the same people who report the news are now writing advertisements, journalistic integrity could be compromised.
This webpage address is called the Uniform Resource Locator, or URL.
Your computer converts the 1’s and 0’s into tiny flashes of light. 1’s might mean to keep the light on for a fraction of a second, while 0’s might mean to keep the light off for the same amount of time. These flashes of light then travel through the fiber-optic cable. Since the fiber-optic cable is made of clear glass,[372] information can travel incredibly fast: about two-thirds the speed of light.[373]
Uber is cheaper than owning a car if you drive less than 9,500 miles a year.
EC2 lets you run your app’s code on Amazon servers,[457] while S3 lets you store all your app’s data on those servers.
Technologists have acronyms for these cloud computing providers as well. AWS, Azure, and Google Cloud Platform are all infrastructure-as-a-service (IaaS; we have no idea how to pronounce it), which lets app makers borrow servers to run their apps on.
What’s the difference between SaaS, IaaS, and PaaS? Let’s use an analogy: food. SaaS is like a restaurant: you just tell the waiter what food you want, and they’ll bring it to you. IaaS is like renting a kitchen: you have the space, but you have to bring your own ingredients and utensils and cook the food yourself. PaaS sits in between SaaS and IaaS: you give someone your ingredients and recipe, and they’ll prepare the food for you.
cloud hosting services like Amazon Web Services will instantly grow or shrink the computing power given to your app as your app’s usage goes up or down during the day, so you only pay for as much usage as you need.[477] This is elasticity.
the cloud builds in lots of “redundancy,” or multiple copies of the same information or code. Even if a few computers fail, there will be plenty of others to pick up the slack.
It took Netflix seven years to completely migrate from their own servers to Amazon Web Services — and, along the way, they basically had to rebuild all their infrastructure and databases from scratch.
one of the biggest pitfalls of cloud computing: if you run your app in the cloud, you’re in trouble whenever your provider has an “outage.”
As Google cofounder Eric Schmidt put it in 2010, “every two days now we create as much information as we did from the dawn of civilization up until 2003.”
Using this data, Target can start finding patterns and making predictions about customer behavior. For instance, Target found that women who suddenly bought large amounts of unscented lotion were likely around the beginning of their second trimester, because these purchases tended to correlate with births a few months later. Pregnant women were also likely to buy supplements like zinc, calcium, and magnesium.
the challenge for retailers like Target is capitalizing on these customer insights without coming across as “creepy.”
It’d be far too expensive to make a supercomputer powerful or big enough to handle all this data itself. Instead, the key is to break your data and computation into more manageable chunks, which you assign to an army of cheap, normal-sized computers.
Amazon can even use the words you highlight on the Kindle to predict what you’re going to buy.[552]
the NSA linked the WannaCry attack to the North Korean government.
people will only pay the ransom if they can trust that the scammers will decrypt their files.[589] This has, strangely, led to scammers having excellent customer support, even sometimes having call centers and online chats with sales representatives.[590] Some even hire designers to make their websites look attractive.[591] They know that they need to build up a reputation of “trust” with their victims
there are many places where malware can sneak in, such as via downloaded files and installed apps.
Some people estimate that the deep web is 500 times larger than the “clear” web that we can find through Google.
Instagram, a platform that originally had no monetization strategy; Instagram ads now pull in over $8 billion a year.[971]
in 2017, Instagram copied Snapchat’s most famous feature, Stories,[973] which demolished Snapchat, slowing the app’s growth by 82%.
In 2018, Facebook announced that its growth had flatlined in the US and Canada and had, in fact, begun shrinking in Europe.
It’s too late to get into China and too early for Africa
Chinese law requires tech companies to hand over user data to the government,
(India has more smartphones than toilets!
In the developing world, financial literacy is a problem, many people don’t have the necessary IDs to start a bank account, banking infrastructure is poor, women aren’t empowered to handle their finances,[1074] and people flat-out don’t trust banks with their money.
There’s even a game where you can send money to a group chat — but only the first member of the group to open the chat gets the money.
As soon as the hongbao feature launched, the number of WeChat Pay users mushroomed from 30 million to 100 million.
