Crashes can happen to everyone and, in the Internet world, there is no such thing as “too big to fail”. In fact, when we talk about server overloads, downtime and crashes, the bigger you are the more risk you have. More traffic means a larger load on your servers and you need to be prepared for it to offer the best user experience possible. The more you grow, the more you’ll need to plan for what’s to come.It also means that sometimes even the most famous companies in the world can have load problems. Their crashes can be taken as examples for other growing companies whose marketing strategies include sads, stunts, or promotional operations and which will lead to an increase in traffic.
Let’s take a look at some recent crashes, identify the issues, the consequences and see what we can learn from them to help you avoid downtime.
Taylor and her last album
The first crash we’re talking about is also the most recent. On October 21, 2022, Taylor Swift released her long awaited album, “Midnights”, and broke Spotify at the same time. More than 8000 outages were reported as the service underwent a lot of stress. Afterwards, it was obvious that Spotify faced a very rare event since the album immediately became the most streamed album in a day in the history of the platform with 228 million streams, while the previous record was 153 million.
What were the consequences of this crash? Spotify works with a subscription model but also with ad revenues. They ultimately lost revenue, as ads weren’t seen, and a piece of trust from their users although for Taylor Swift it was a very nice buzz.
Spotify wasn’t alone as Ticketmaster also faced large issues when they opened sales for the Era Tours, the tour following the release. With over 3.5 millions shoppers queuing to book a ticket, the platform, even though it verifies people, excludes bots and is accustomed to large events, broke all its personal records. With people losing booked tickets and occasional downtimes, the company faced resentment from the fans and even the star herself. No need to explain where the harm lies here.
Coinbase broke during Super Bowl
Another famous website crashed in 2022. It occurred during the Super Bowl, the event of superlatives, viewed in 2022 by almost 100 million people, with famous ads,a half time show and everything in between. So, when you’re a brand taking part in that kind of massive event, you have to be prepared.
Coinbase did their math nearly perfectly since the downtime was only minutes long but unfortunately, it happened when their ad got screen time during the Super Bowl. A QR code leading to a promotional page is a great way to bring people to your platform, but it comes with a risk depending on where you show it. And in fact, their landing page got hit by 20M clicks in one minute. Their payment process also showed weaknesses during press time.
It could have become very complicated for the company if it turned to be an extended crash or slow down. In the trading industry, time is of the essence and downtimes are not appreciated. Initially, trading businesses tended to have their offices as close as possible to the stock market building in order to have the lowest latency when placing orders. You need to be able to react quickly, if not instantly, to the market evolution, not even considering the possibility of downtime. If your users start to grow doubts about your technical capabilities to offer a solid trading platform, you might be in danger. Especially in this case where it could be the first experience for any of these prospective users.
Pokemon Go early days
A few years back, in 2016, Niantic took the world by surprise with one of the most addictive games ever created, Pokemon Go. Cities were swarmed with Pokemon catchers looking frantically at their phones… at least when they could connect to the platform. In fact, as the player base grew, the game got hit with overload issues on the first weekend after its release.
Some might say that every online video game experiences that kind of problem during the launch phase, and that’s broadly true. The server allocation is tricky and that’s why online games go with open betas and stress tests to prepare prior to launch. It’s not rare either to see MMORPGs opening new servers in a hurry a few days after a launch once they better understand their traffic and requirements.
However, in the case of Pokemon Go, it crashed on the first weekend. Fortunately, Niantic managed to fix it quickly and before long the application managed to stay stable for a long time. Not bad for a game which had 250 millions players at its peak.
Since the initial crash, Pokemon Go, has experienced additional issues during times of high use such as the incident we detailed in our Load Testing for Mobile Gaming blog post.
The long-term damage to Niantec wasn’t that serious but it still reminds everyone that you need to prepare for success. No matter how resilient your environment is, if you are looking for growth you will need to balance your current and future needs, thus asserting what you can handle at every moment. Regularly load testing for scalability is key.
Chipotle went wild with free guac
Faith can move mountains and apparently so can free guacamole. Chipotle made that discovery in 2018 during National Avocado Day when they offered free guacamole for mobile and online orders. The result? Their servers crashed.
This is a case study of the worst that can happen when a website or application crashes. You plan a promotion, for a dedicated amount of time and for food, which means the demand window will be very short, aka lunch or dinner time. You will create a traffic peak of shoppers your servers will have to handle. What happens if it crashes with a high demand and a very desirable product? You generate frustration, a lot of it, as well as a lot of cash loss.
In the end, Chipotle had to extend the offer for a day in order to compensate for the failure. To add a final touch, this issue happened at a time where the brand was under pressure from food-safety issues. Remember that promotional actions can come at a high price if you can’t deliver them.
Amazon shut down for an hour during Prime Day
Amazon ran into a similar situation in 2018 during Prime Day. hoppers were blocked for a few hours but the operation still became their best Prime Day at the time. The difference with Chipotle was that their demand window was far larger so the backlash was far less severe.
The cause was simple, the servers couldn’t handle the load and they had to make a few technical adjustments to keep the website running, such as manually adding servers or blocking international traffic for some time.
The irony here is that AWS is also one of the world’s largest cloud providers, meaning they had the resources to prepare. The same Amazon found in 2016 that a 100ms latency would cost them 1% in sales. What loss would they calculate with a full crash then? And what could they have done better to handle Prime Day? Better server allocation or request volume anticipation? A bit of both probably, now that they have a better idea of the heights their traffic can reach.
An overload can happen in any industry, from IT to Business Services and even Food. In the end, it’s only your incoming traffic against your servers and their capacity to handle it properly to provide the best user experience.
Avoid the pitfall of thinking that because it works now, problems won’t appear in the future. A company that strives to develop, expand, grow, will also increase its traffic as it does so. As you saw in this article, crashes, latency and downtimes happen to famous businesses for different reasons and they always have consequences for them. You can’t plan everything regarding traffic, but you can’t stay idle either.
That’s where load testing comes into play. By load testing before launch and continuing to test in your development cycle you can find your bottlenecks and weaknesses so you can correct and adjust before it’s too late.
Gatling can provide the information you need to prepare your website and make sure that it won’t fail you during important events, even if it’s not Prime or Avocado Day.
Let us know about your project
Used by over 100,000 web applications and e-commerce websites, Gatling solution is put to use in diverse industries including Software, Streaming media, Banking, Insurance, Gaming and Finance.
Taylor Swift’s ‘Midnights’ album crashes Spotify, leaving fans shocked; nearly 8,000 outages reported – Fox Business
TAYLOR SWIFT | THE ERAS TOUR ONSALE EXPLAINED – Ticketmaster
Most day-one streams of an album on Spotify (male) – Guinness World Records
Coinbase Site Crashes From Traffic as Super Bowl Ads Spark Public Interest – Blockworks
Pokémon Go makers call for calm as servers crash across Europe and US – The Guardian
Chipotle’s ‘guac gaffe’ crashes mobile app, website – Marketing Dive
Why Amazon’s Site Crashed on Prime Day – Fortune