Friday, May 26, 2017

How video impacts mobile web performance and UX, part 3: detecting and remedying issues

Video has become an important tool in the marketers’ tool box. Video storytelling is a useful and increasingly popular way to engage customers.

But if your video doesn’t work properly or cripples your website or app performance it will become a major frustration to customers, marketers and techies alike.

In the previous two parts of this column on data and download speed and autoplay and audio we learned:

  • Video dominates mobile data traffic
  • When implemented correctly, mobile video should not impact the speed that pages load on a mobile device
  • Mobile users start to become impatient after waiting just two seconds for a video to load; by 10 seconds a fifth will have given up.

This column will explore how to detect, avoid and remedy issues with videos to give your viewers the best possible experience with your video content and keep them engaged and watching your videos.

Jump to:

How to detect problems with video
How to avoid problems with video
How to remedy problems with video

How to detect problems with video

Detecting issues with video, audio or any other web or app issue a) can be straightforward; b) should be everyone’s responsibility, from the CEO down; and c) helps to keep agencies, techies and marketers on their toes.

1. Use it

Blatantly obvious – but when was the last time you checked out your site and videos from a bus, train or bar? Incentivize employees to use the site/app (during beta testing and routinely after goes live) and report issues and suggest improvements.

Check for:

  • How quickly did the site/page load? (Count the seconds)
  • How long did you have to wait for the video to start?
  • How good is the quality?
  • Does it stall / (re) buffer during playback?
  • Was it worth watching/watching to the end?
  • How do you feel about these conclusions?

2. User test it

Recruit customers and monitor their behavior and reactions as they use your web site, using different devices, networks and locations. Score against the above checklist. If this cannot be conducted in person use a remote service such as UserTesting.com.

User testing should occur at each stage of the development process. For more on why user testing is so crucial, see my previous column for our sister site ClickZ on Why user testing should be at the forefront of mobile development.

3. Test it

There are different types of testing, including:

  • Page performance – tools such as WebPageTest (free) show how/if the video is impacting how fast the page loads. It shouldn’t. The image below shows the WebPageTest results for how quickly Sam Dutton’s mobile video explainer on YouTube loads on a mobile device. The page took 6.6 seconds to load 809kB.
  • A/B testing – tests alternative experiences with different groups of web (or app) visitors. For example, test hosting the video on the homepage versus on a dedicated page.
  • Video testing tools – AT&T’s Video Optimizer (formerly known as Application Resource Optimizer) is a free-to-download tool used by developers (requires technical knowledge) to detect issues such as delays with start-up and the frequency and duration of stalls and optimum segment size.

Image shows the WebPageTest results for how quickly Sam Dutton’s YouTube video loads on a mobile device. It took 6.6 seconds to load 809kB.

4. Monitor it

  • Web analytics tools, such as Google Analytics, track visitor engagement with video – e.g. number of views, who viewed, how long, and with the webpage itself, including dwell time and bounce rate. See this introduction to using GA to assess video engagement.
  • Heat map tools, such as Clicktale and Crazyegg provide a visual representation of how users interact, or attempt to interact, with webpages and video.

How to avoid problems with video

Following best practices while creating/producing the video or coding the page, website or app that will host it should help avoid many of the common issues – videos that won’t play, are slow to play, or have broken playback.

Industry guidelines on mobile video are thin on the ground, considering the increasing popularity of the format. What guidance is available tends to be a bit techie and thus a turn off for non-techies.

The following recommendations have been compiled with the help of:

1. Make it worth it

There are many costs involved with video/audio:

  • For the producer: the cost of production and distribution; impact on web performance
  • For the network: the impact of network congestion
  • For the viewer, in terms of data consumption, battery life and time it takes to consume.

This makes it imperative that the video is meeting a known user need, contains quality content, is the right length, optimized in terms of bitrate, segments and compression.

2. Be aware: video is greedy; HD greedier; 4K much greedier

When it comes to bandwidth, standard video is greedy, requiring 0.5 Megabits per second (Mbps); high definition (HD) is five times as greedy as SD; and 4K is 30 times as greedy.

Cisco’s Usha Andra explains:

“Mobile video and multimedia applications have higher bandwidth and lower latency requirements than non-video applications. The requirements can range from a low of 0.5Mbps for standard definition (SD) to 2.5Mbps for high definition (HD) and over 15Mbps for 4K/ultra-high definition (UHD) downloads and much higher for virtual reality (VR). Latency requirements can range from 100 milliseconds (ms) to 15ms for UHD VR video applications.”

3. Know the limitations of mobile networks in your target markets

Even among developed telecoms markets, the capability of mobile networks varies considerably. Check the Cisco GCI Global Cloud Readiness Tool for an averages of each country.

The stats suggest that download speeds in the US and UK are 40% lower than Norway and South Korea, and 25% lower than Canada:

  • South Korea – download: 31.0Mbps; upload: 14.3Mbps; latency: 68ms
  • Norway – download: 29.1Mbps; upload: 11.6Mbps; latency: 40ms
  • Canada – download: 24.2Mbps; upload: 9.0Mbps; latency: 51ms
  • UK – download: 18.2Mbps; upload: 8.0Mbps; latency: 55ms
  • US – download: 17.1Mbps; upload: 10.0Mbps; latency: 88ms.

Usha Andra adds:

“Please note that these are average speeds and latencies, which means many users experience higher or lower speeds compared to the average speeds. When the speeds and latencies are lower than what an application warrants, the end user experiences delay in video, garbled audio, etc.”

4. Home page or own page?

Few of the most popular sites, including those that have a strong video focus – YouTube, Vimeo, BBC and CNN – host videos on the homepage or category pages. These sites promote their videos on the homepage as image links (often with play button icon overlaid) and text links, which when clicked or tapped go to a page dedicated to that video.

Why? Keeping video off the homepage keeps it leaner and faster to load on mobile devices. See the Twitch example below.

5. Avoid autoplay

Forcing mobile web visitors to view video whether they want to or not, is:

  • Frustrating for the customer (especially when it happens in a quiet environment)
  • Prone to using up the customer’s bandwidth and battery life unnecessarily
  • Liable to slow down how quickly the page loads
  • Contrary to accessibility best practice (as it can interfere with the screen readers used by visually impaired people)
  • A common technique for artificially inflating video view stats.

There is a (vaguely plausible) argument that sites such as YouTube are an exception to the no autoplay rule. As the visitor is clicking through to the video on a dedicated page it is implicit that they intend to watch.

Consider Twitch, the surprisingly popular site where fans watch gamers playing video games live, captured in the image below. On the desktop homepage, Twitch.tv has a live game on autoplay, while on m.Twitch.tv, there are no videos hosted on the homepage.

Comparing the download size and page speed of Twitch homepage when downloaded to a mobile and desktop device on HTTP Archive (April 15 2017) delivers dramatic results:

  • Mobile homepage (with no video) took 5.8 seconds to load 354kB of data over 24 requests
  • Desktop homepage took 19.9 seconds to load 16,255kB of data over 275 requests. Of that, 11,827kB is video content.

Image shows the desktop homepage and mobile homepage of Twitch.tv. Desktop has a live game on autoplay, but there are no videos hosted on the mobile homepage.

6. Viewer experience (VX) and choice

Make sure the video and host page is intuitive. Let the viewer take control. Make it easy to:

  • Choose video quality – low quality, HD or 4K
  • Select and exit full-screen view
  • Change device orientation change
  • View and operate. Ensure the video fits the device screen and that buttons are intuitive
  • Allow playback when the device is offline.

7. Make the video accessible

To make video/audio accessible for:

  • Visually impaired people, provide a written transcript of the audio.
  • People with hearing impairments, provide subtitles.

For more advice on making mobile content accessible to a wide audience, the BBC Mobile Accessibility Guidelines are an excellent resource.

8. Minimize video start-up delay

The delay to start-up is caused by two essential processes:

  1. The authentication process (including digital rights management).
  2. The downloading of the video. Video files are subdivided into segments. A sufficient number of segments need to be downloaded to the buffer (temporary store on the client device), before the video starts to play.

A delay is inevitable, but the video should be optimized to ensure delays are kept to a minimum.

As can be seen from the 2016 data from Conviva study below, videos tend to take longer to start on mobile devices, both on WIFI and Cellular, than Tablet or Desktop. It’s no coincidence that mobile has the highest proportion of exits per attempt.TAG: First graph shows video start up times for devices including Desktop 3.1 seconds, tablet cellular 2.9 seconds, mobile cellular 3.7 seconds, mobile WIFI 3.5 seconds. First graph shows number of exits as percentage of attempts for same devices including Desktop 12%, tablet cellular 16%, mobile cellular 19%, mobile WIFI 17%.

9. Keep the user informed

While the authentication, downloading and (re) buffering occurs, tell the user what is happening and/or distract them. Watching a spinning wheel icon can be frustrating.

10. Minimize video stalls

Stalls occur when too few video segments stored in the buffer to allow playback to continue. The video will not continue until sufficient segments have been downloaded (called re-buffering).

The key is to find balance between slow start and stalling, says AT&T’s Doug Sillars:

“The 2 biggest metrics for video are:

  1. Startup delay (how long from click to stream).
  2. Stalls (video stops, maybe a spinner).

These are (of course) interrelated.  If you startup too quickly – there will not be enough video stored locally on the device… and you might get a stall.  Or you can take too much data at the start (long startup delay), but have no stalls later.

There is a magic “Goldilocks” point in the middle – not too hot, not too cold – that balances the two factors.” 

11. Optimize bitrate, compression and segment size

Optimize bitrate, compression and segment size for the device and network connection.

  • Re-buffering typically occurs where the video is played at a speed, measured in bitrates (bits per second), that is too fast for the download speed (bitrate) of the network connection, so the buffer is emptied quicker than it is being filled.
  • Digital videos are divided into files, called segments, of 2 to 10 seconds, which are downloaded to the buffer and then played in order. Segments of optimum size for the connection will download, buffer and play faster.
  • A Codec (coder/decoder) is a tool for compressing and decompressing audio and video files. There are a number of different compression formats, e.g. MPEG-4, each with pros and cons. Different video quality and the client device/connection will influence choice of format.

12. Use adaptive bitrates.

Adaptive bitrate streaming creates and stores digital video at a number of different quality/speeds/bitrates. The video player on the client device requests the most appropriate of these based on a) network speed, b) device capability, and c) capacity of the buffer.

There are two types of adaptive streaming, DASH and HLS, because one industry standard that worked on all devices would be just too easy (find out more here).

13. Use a content delivery network (CDN)

A content delivery network speeds up how quickly web media, including video loads and plays on a mobile device by reducing the that the video has to travel between the original web server – e.g. your webserver in California, USA and a viewer in Timbuktu in Mali – by replicating and storing the video on servers around the world.

According to BuiltWith, 53.8% of the top 10k websites use CDNs.

Akamai Edge, which was one of the original CDNs, founded in 1999, remains one of the most popular. According to BuiltWith, Akamai is used by 11.4% of the top 10,000 sites, followed by Amazon CloudFront at 4% and MaxCDN at 1.3%.

Graph shows the market share of CDN providers according to BuiltWith: Akamai with 58%, Amazon with 20%, MaxCDN with 7%, EdgeCast with 6%.

14. Host or embed?

Hosting websites on a third party network, and embedding the file, removes several headaches, including video compression, adaptive bitrates and engaging a CDN. This helps to explain why 15.2% of top 10k websites embed YouTube videos and 3.6% Vimeo, according to BuiltWith.

How to remedy problems with video / audio

1. Page weight or load speed issues.

Regularly check the key pages using a testing tool such as WebPageTest (this is the tool used by HTTP Archive).

If this highlights issues of excessive page weight, slow download speed, and it appears that video is a contributing factor (rather than oversized images or inefficient use of JavaScript), the options are:

  • Kill autoplay
  • Ensure the video is not preventing the page loading correctly
  • Move the video to a dedicated page (with a prominent picture and text link)
  • Use A/B testing to verify if this solves the issue.

2. Video fails or is slow to start or stalls during play

If the video performance is an issue, here are some troubleshooting tips to try:

  • Try loading the video to a dedicated video service such as Vimeo or YouTube. Compare the performance of the video on the third-party site, embedded on your site and with the self-hosted version to highlight if problems lie with the video, as opposed to the website, webserver or CDN (or lack of CDN)
  • Test the video with a tool such as AT&T’s Video Optimizer (requires development skills) to detect issues with video segmentation, compression, buffering etc. and fix them
  • Have the video re-edited to make it more concise; and optimized to improve bitrate and compression
  • Use or replace the CDN.

If video performs better on some devices and over different connections e.g. PC on cable versus smartphone on 3G:

  • Prepare a number of versions of the video in different formats, with different quality, bitrates and compression to suit the most common scenarios of device and network type
  • Use device detection to discover the client device, its capabilities and the type of connection to serve the most appropriate version of the video
  • Use adaptive bitrates.

Resources (and sources)

These resources are aimed at developers, but are useful for all (if you ignore the techie bits):

 

This is Part 3 of a series looking at how video impacts mobile web performance and UX. Read the previous installments:

Thursday, May 25, 2017

Taming the local search beast in a post-Possum and Fred world

It’s estimated that 46 percent of all searches performed on Google have a local intent, and the Map Pack appears for 93 percent of these.

In September 2016 Google unveiled a new local search algorithm, dubbed Possum, and it pretty much went unnoticed in comparison to the real-time Penguin update released in the same month.

In short, Possum made it harder for businesses to fake being in locations that they’re not (through the likes of virtual offices), as well as tackling Google My Business spam.

Possum, however, isn’t a “single” algorithm update, as it affected both localized search results as well as the Map Pack, which of course are two separate algorithms both triggered by search queries that are interpreted as having a local search intent.

The Google “Fred” update, which hit SERPs back in March, has also had an impact on local search, much like the Phantom updates before it.

A lot of local SERPs are extremely spammy, where websites have been built cheap and location names have been liberally applied to every menu link and keyword on the page, such as this home page sidebar menu:

This of course, is only a snapshot of the page – the menu and tile icons go on a lot more. Spam such as this still ranks on page one, because Google still has to provide results to its users.

Take advantage of the market conditions

A lot of locally-focused websites aren’t built by agencies; the vast majority tend to be self-built or built by bedroom level developers who can churn out a full website for £300 (or less).

Some verticals have seen some significant online investment in recent years, while others lag behind considerably. By investing in a good website and avoiding the same spammy tactics of your competitors, you can create a powerful resource offering user value that Google will appreciate.

Directory submissions and citations

Just to be clear, I’m not talking about just backlinks. Recent studies have shown that citations with a consistent NAP (Name, Address & Phone number) are important to both local algorithms.

There is no magic number to how many directory submissions you should have, but they need to be relevant.

I’ve worked on local campaigns in the UK where they have been previously submitted to directories in Vietnam, Thailand and Australia. Yes, it’s a backlink, but it’s not relevant in the slightest.

Think local with your directories, and exhaust those before moving onto national ones. The number of local directories should also outweigh the nationals were possible. To do this properly, it’s a manual process and to ensure quality it can’t be automated.

Reviews

Review volume, velocity and diversity factors are important, and in my opinion, they’re going to become more important in the coming months – particularly following the recent release of verified customer reviews for online businesses.

In Google’s Search Quality Evaluator Guidelines, the evaluators are instructed to research a website/brand’s online reputation from external sources in order to assess the quality of the website.

This is why getting reviews on your Google My Business listing, Facebook pages, positive tweets, Yell, Trip Advisor reviews etc are all great. Having testimonials and reviews on your website is great for users, but you wouldn’t publish bad reviews on your own website, would you?

Google accepts that negative reviews appear, but as long the good outweighs the bad, you shouldn’t have anything to worry about. If you do get a negative review, demonstrate your customer service and respond to it. You can set up Google Alerts to monitor for your brand and flag up any external reviews.

Screenshot of Amazon reviews for a product, averaging 4.7 of 5 stars.

Google My Business & Bing Places

Believe it or not, Google My Business is considered a directory, as is Bing Places. It’s important that you have one if you’re a local business, and that you’ve optimised it correctly. This means the correct business name, address and phone number (keep your NAP as consistent as possible), choose an appropriate category and include a thorough description.

localBusiness structured data mark-up

Structured data mark-up (or schema) is an addition to a website’s code that enables Google’s RankBrain (and other AI algorithms from other search engines) to better understand a website’s context by providing it with additional information.

Not all websites are currently utilizing this schema (or any schema), and Google wants you to use it.

If you don’t have developer resource to hand, and you’re not a coder you can use Google’s Data Highlighter to mark-up content – you will need a verified Google Search Console however to make this work.

Other considerations

As well as focusing locally, you need to also consider other ranking factors such as SERP click-through rates.

Optimizing your meta title and description to appeal to local users can have a huge impact on click-through rates, and the change could be as simple as including the phone number in the title tag.

You also need to be on https and have a secure website. Getting hacked, suffering a SQL injection or having malware put on your site can seriously damage your reputation within Google and take a long, long time to recover.

Wednesday, May 24, 2017

What does Google’s “Project Owl” mean for search and fake news?

Have you heard of Google’s “Project Owl” yet?

If not, then you’re in for some fun, because this is a hoot.

Let’s start at the beginning.

Fall of 2016: Trump gets nominated to the presidency.

Still in fall of 2016: All around the world, people are asking “WHO? WHAT? HOW?” That’s when researchers found that American voters were influenced by misinformation on the internet.

The world is completely distressed. They demand that 1) someone be responsible and 2) for them to take action and fix the ‘fake news’ problem.

Oh – hi Larry Page and Mark Zuckerberg.

Who else other than Google and Facebook, right?

The public wants solutions from search engines and social media giants to tackle ‘fake news’ and any other misinformation on the internet.

May 2017: TA-DA! Welcome Project Owl.

Project Owl is introduced as Google’s answer to addressing fake news. It plans to do this with new feedback forms for search suggestions and the answer box, and authoritative content prioritization in the answer box.

And no, we don’t see this affecting marketers or SEOs. As long as you continue to practice white hat methods, your day-to-day should be the same. However, given this can affect searchers’ user experience, we see a few challenges.

Challenge #1: Search engines are supposed to be neutral

Google is walking on a tight rope. If search engines manage to accomplish tackling fake news, then first, that feels like a violation of the first amendment but second, they will come off as bias to specific news/media sources.

Remember, feedback from some users will change the search experience for all on that query. It will be difficult to differentiate what’s ‘right’ for one searcher versus what’s ‘right’ for the other.

But, you know what? When personalized search engines are the new thing, this may not even be a challenge.

Challenge #2: The proposed plan

Let’s take a step back and look at Google’s track record when they are “working to fix” something. Just like many updates in the past, Google says one thing and marketers notice something completely different.

Right now, “Project Owl”, according to Google, will rely on the searcher to provide feedback on the autocomplete or on the featured snippet.

But, we’re missing the obvious.

Let me ask you: When was the last time you went in and changed any of your Google search settings? Or rather, did you even know that it was possible to change Google search settings?

Don’t feel bad – I know SEOs who didn’t even know they could do that!

Google said and I quote, “We plan to use this feedback to help improve our algorithms.” That is what they told us years ago about link disavow and they still don’t have that right. My take is that it will be several years before Google is able to filter out “fake news”.

I personally think TMZ.com spreads lots of fake news, yet they rank for 2,133,648 keywords on Google; and I don’t think Google is going to start taking their keywords away anytime soon.

As you can see I don’t think Google is going to put much into this and even if they do it will take years before it’s perfected. I believe Google is in crisis mode right now but sooner than later people will forget and Google will move on or deprioritize this.

Challenge #3: Obscure and infrequent queries

The third part of Google’s solution is prioritizing authoritative content specifically for obscure and infrequent queries. But, when it’s already such a niche group, how can you determine who that authority should go to?

Challenge #4: The blackhats

Like every other SEO tactic, there is always the one group of SEOs that capitalize on Google making an algorithmic change or giving us the capabilities to affect how the algorithm reacts.

I know blackhat SEOs are going to jump at this chance to devalue other people’s content that don’t serve theirs or their client’s interest. They will probably work from C class IP addresses and run bots on specific timing intervals to make it seem natural.

Now what?

Overall, a first step is better than no step at all, but here are two ways I recommend as a stronger combat against fake news.

First, Google should not only rely on end users to report content that is fake or offensive. Its focus should be less on that and more on perfecting RankBrain, Google’s artificial intelligence.

Second, it’s not just up to the Googles and the Facebooks to take action. It’s also a user’s responsibility to determine whether a search listing is worthy of your click and trust.

When you see something that sounds outrageous, it probably is. Hoaxes appeal to natural human curiosity, which is why it’s hard not to click, but still, that’s a choice you get to make.

Tuesday, May 23, 2017

Targeting generational buzzwords like “Millennials” means targeting no-one

If I were to tell you that marketers were using astrological signs as a way to understand/target specific groups of people, you’d tell me that’s a ridiculous strategy.

“Astrology is fake,” you’d say, and given the precision of modern marketing tools, using the stars to analyze customers or understand population segments would not only be lazy, but the chances of it working would be random at best. Yet, this is happening daily.

How? For example, thinking that millennials, a 75.4 million cohort of people in the United States alone, share a universal set of attributes.

Speaking in absolutes about a demographic that makes up ~20% of the total population of the United States with nearly no shared characteristics completely ignores the nuance, depth and uniqueness of humanity, and our diverse wants, needs and desires. We are complex creatures!

Common sense would indicate that drawing conclusions about such a loosely defined group of folks is at best “pushing it,” and at worst completely ludicrous. There’s simply no way to make an accurate, universally applicable statement about that many people, defined solely by a 20+ year age range based on the year they were born.

There’s no rigorous methodology behind generational branding

Even if I wanted to take generational branding seriously, it’s in my opinion not good social science. “Baby Boomers” (18 year cohort) are defined as people born between 1946 – 1964, and an age range between 51 and 70.

Millennials” (a 23 year cohort) are people born between 1981-2004, giving an age range of 12-35. Gen Z (no defined cohort yet) have birth years that range from the mid-1990s to 2000s, and, so far there is little consensus about ending birth years.

The ranges are not only inconsistent, but the fact that not everyone can even agree on these unstandardized, randomly assigned dates says it all. It’s all highly questionable, even for a softer science like sociology.

A ~20 year ago cohort is too large to mean anything when our experiences of media, culture, etc. have fragmented

Social trends now move so quickly that single moments of significance are less defining, even if at the time they were seemingly important. The 3-TV-channel world where we all watched the same things has been dead for decades and yet we still apply concepts that were created then.

Everyone’s experience of the world from a media perspective alone is so unique we can’t underestimate the number of niche communities that now exist that have less to do with age and more to do with personality. The world and the people in it are becoming more, not less, complex and we need updated thinking if we hope to understand it and market to it.

Psychographics show far more in common than year born / demographic breakdown by year born. If you can target, not just arbitrary ranges as defined by buzzwords, but by people who live in a specific area, are married and are interested in weightlifting and organic food you would have to be willfully ignorant or lazy to think stepping back and targeting everyone is a good idea.

With the depth we have available for ad targeting in tools like Google AdWords and Facebook ads, it’s inexcusable to not take the time to target the right message to the right users. The sophistication of our marketing capabilities means we’re doing our shareholders and customers a disservice not to go deeper.

Sample AdWords ad targeting capabilities mean reaching specific and precise segments relevant to us:

Sample Facebook ad targeting capabilities reach specific social communities that care about our brand:

As for marketing to specific age ranges? Of course there are product categories with immutable segments for a certain demographic. But buzzwords like “Baby Boomer” aren’t required to market to these groups effectively.

Additionally, you want to be more specific than a 20 year cohort to accomplish this in a meaningful way. For example, a 34 year old millennial living in a city has little in common with a 20-something millennial just finishing college in a small town – yet generational buzzwords lump them together.

In Google Analytics, we break out age ranges in smaller, more manageable chunks, so you can analyze college-age students in a specific area which would be far more instructive.

To some, the word millennials has become just a blanket term for young people. This almost comical story of an iconic American brand grasping for relevancy shows what may be a typical situation in boardrooms, where a group of executives clearly feels behind the times.

So it seems like an easy solution to just use broad strokes like buzzwords. A brief quote from this story illustrates:

The other challenge is that many people who work at American Express aren’t all that millennially minded themselves. If you visit Amex’s headquarters in Lower Manhattan, you’ll find squared-jawed men in bespoke suits and fashion model-glamorous women, but not a lot of young people in the uppermost ranks … In one Amex brainstorming session, according to an executive I spoke with, participants spent 10 minutes trying to figure out what FOMO meant before turning to Google. They discovered it stands for “fear of missing out.” It is unclear if the group recognized the irony.

I don’t think this habit of over-generalization comes from a desire to marginalize millennials, but I do believe it’s a broader way people use to try and make sense of a technology-driven world.

In most analyses of millennials, the way technology shapes and controls their environment is key to understanding whatever point is being made about them. This categorization provides a way to add a human layer to the discussion around those who have been born into a world where technology and the internet automate our existence.

Why waste time with generational buzzwords when we have so many better groups to analyze/target/study instead?

For example, with recommended actions: 

  • Users who responded to holiday ads last year that become recurring customers over the next year (run more of those specific ads next season, replicate for your other product categories and double the budget if the numbers were previously great!).
  • The specific location with the highest average purchase order or customer loyalty for a national restaurant chain (or better yet, the top 5%). What went right here? What are the common traits among customers here and how can we attract more of them to our other locations?
  • For a pharmaceutical company with a new arthritis drug, targeting people ages of between 30 and 60, the average onset of RA According to the Arthritis Foundation (this is a specific, actionable age segment, not the nebulous “baby boomer” and is immutable range, no buzzword required).
  • All your site visitors who added something to their shopping cart but don’t complete checkout. For sure these include people of all ages; likely optimizations don’t even require demographic data.
  • Users who follow your brand on social channels (aka your influencers) – what can you learn about this very specific group that is unique to your brand. Incredibly useful to understand these folk and their nuances so you can best nurture those relationships.
  • The top 20% of your customers by annual spending or product category. How can you grow these really valuable segments?

The above list is just to get you thinking, but to me it’s so exciting what’s now possible that to keep doing what was always done is doing our work and sector a huge disservice.

Or, you could just ignore all of this and just make stereotypical ads for millennials without actually getting to know them, so that you too can repeat Pepsi’s gaffe and become a global embarrassment.

Google I/O: What’s going on with Progressive Web Apps?

At Google’s developer jamboree, Google I/O, last week the search giant paraded a host of big name case studies and compelling stats to herald its success with two initiatives to make the mobile web better and faster: Progressive Web Apps (PWA) and Accelerated Mobile Pages (AMP).

Progressive Web Apps are a Google innovation designed to combine the best features of mobile apps and the mobile web: speed, app-like interaction, offline usage, and no need to download anything.

Google spotlighted this relatively new web product at last year’s Google I/O, where the Washington Post showed off a newly-built Progressive Web App to enhance its mobile experience.

Whether companies believe in or plan to adopt Progressive Web Apps, the initiative (along with AMP) has done a fantastic job of highlighting a) the importance of making websites and apps lean and mean so they perform better on mobile and b) how ridiculously bloated, slow and inefficient websites and apps have become.

PWA and AMP are not the only answers to mobile bloat, but being led and backed by Google, they bring the potential for 1) broad adoption, 2) lots of resources, and 3) favorable treatment from Android, Chrome and Google Search.

What’s so great about Progressive Web Apps?

PWAs bring native app-like functions and features to websites. They should (depending on the quality of the build) work on all smart devices, adapting the performance to the ability of the device, browser and connection.

The key features that get people excited about PWAs are:

  • The ability to send push notifications
  • Option to save to the device (home screen and – now – app launcher), so it loads even faster next time
  • Ability to work offline (when there is no internet connection)
  • Make payments. One of the most significant PWA announcements at Google I/O was that PWAs can now integrate with native/web payment apps, to allow one tap payment with the users preferred provider, including Android Pay, Samsung Pay, Alipay and PayPal
  • Closer integration with device functions and native apps.

The margin of what native apps can do compared with a web-based app (N.B. PWAs do not have a monopoly over mobile web apps) is disappearing rapidly.

The last year has seen a remarkable 215 new APIs, allowing web apps to access even more of the native phone features and apps, announced Rahul Roy-Chowdhury, VP, product management at Google, in his Mobile Web State of Union keynote.

He pointed out that you could even build a web-based virtual reality (VR) app (if you wanted to), citing Within and Sketchfab, which showcase creations from developers around the world.

Who ate all the pies?

But the most compelling thing about Progressive Web Apps is their download size, compared with iOS apps and Android apps. Check out the size comparisons in the image below for two case studies featured at Google I/O: Twitter Lite and Ola Cabs (the biggest cab service in India, delivering 1 million rides per day).

  • Size of Twitter’s Android app 23MB+; iOS app 100MB+; Twitter Lite PWA 0.6MB.
  • Size of OLA Cabs Android app 60MB; iOS app 100MB; PWA 0.2MB.

Why does size matter? Performance on the web is all about speed. The smaller the size the quicker the download. Think SUV versus Grand Prix motorbike in rush hour traffic.

Image: Who ate all the pies? Size of Twitter’s Android app 23MB+; iOS app 100MB+; PWA 0.6MB. Size of OLA Cabs Android app 60MB; iOS app 100MB; PWA 0.2MB.

Interestingly, Twitter markets the PWA as Twitter Lite particularly targeted at people in tier two markets where connections may be inferior, data more expensive and smartphones less advanced; while Ola Cabs markets the PWA at second or third tier cites where there are similar issues with connections and smartphones.

This (cleverly) helps to position the PWA as non-competitive to their native apps.

Which companies have launched Progressive Web Apps?

A growing number of big name brands (see image below) have launched PWAs. These include:

  • Travel companies: Expedia, Trivago, Tui, AirFrance, Wego
  • Publishers: Forbes, Infobae, Washington Post, FT, Guardian, Independent, Weather Company
  • E-commerce companies: Fandango, Rakuten, Alibaba, Lancôme, Flipkart
  • Formerly native app-only companies: Lyft, Ola Cabs.

Map shows companies that have launched progressive web apps, including Expedia, Trivago, Tui, AirFrance, Wego, Forbes, Infobae, Washington Post, FT, Guardian, Independent, Weather Company, Fandango, Rakuten, Alibaba, Lancôme, Flipkart, Lyft and Ola Cabs.

At I/O, Google trumpeted the achievements of a number of companies, inviting several to share their experiences with the audiences – only the good stuff, clearly.

1. Faster speeds; higher engagement

m.Forbes.com has seen user engagement double since launch of its PWA in March (according to Google).

For the inside track see this Forbes article. The publisher claims its pages load in 0.8 seconds on a mobile device. The publisher was aiming for a Snapchat or Instagram-like experience with streams of related content along with app-like features such as gesture-based navigation.

In this video case study, embedded below, created for I/O, Forbes claims to have achieved a 43% increase in sessions per user and 20% increase in ad viewability.

The Ola Cabs PWA takes 1-3 seconds to load on the first visit – depending on the network, “including low 3G” Dipika Kapadia, head of consumer web products at Ola, told I/O attendees. On subsequent visits it takes less than a second as it only needs to download the real-time information, including cab availability.

Ola achieves this partly due to its size: the app is just 0.5MB of which only 0.2MB is application data. As it downloads it prioritizes essential information, while other assets download in the background.

2. Consumers readily download PWAs to their home screens

When mobile visitors are using the mobile app, they receive a prompt to save it to the home screen, so it loads faster next time. It does this by caching all the static parts of the site, so next time it only needs to fetch what has changed.

Twitter Lite, as Patrick Traughber, product manager atTwitter, told the Google I/O crowd, sees 1 million daily visits from the homepage icon.

Since launch of the Progressive Web App, in April 2017, Twitter has seen a 65% increase in pages per session and 75% increase in tweets.

3. Notifications

The ability to send notifications to mobile users to encourage them back to the app, used to be one of the big advantages of native apps over mobile web. No longer.

Notifying users about recent activity is very important to Twitter, said Traughber. And Twitter is taking full advantage of this capability, sending 10 million push notifications each day.

For the inside track on Twitter’s PWA, see this article.

4. Winning back customers that have deleted your native app

App-only companies face the challenge that users only download and retain a limited number of apps on their smartphone and will uninstall those that aren’t used as regularly as others, thus once deleted, it’s over.

Thus it is an eye-opener that 20% of Ola PWA bookings come from users who have previously uninstalled the native app.

See Google’s case study on Ola’s PWA.

5. PWAs appeal to iOS users

Compared with other mobile browsers such as Chrome, Edge, Opera and Samsung, the default browser on Apple devices, Safari, can be slower when it comes to adopting advancements in mobile web. This means Safari users won’t experience some of the more advanced features of PWAs, yet.

Despite this, brands are seeing improved mobile engagement after launching a PWA. Lancôme Paris has seen session length improve by 53% among iOS users, according to this case study of the Lancôme PWA, cited at Google I/O.

6. Conversions

According to Wego’s video case study, embedded below, created for I/O, the Singapore-based travel service has combined both PWA and AMP to achieve a load time for new users is 1.6 seconds and 1 second for returning customers. This has helped to increase site visits by 26%, reduce bounce rates by 20% and increase conversions by 95%, since launch.

If you need more impressive stats to make the case for a web app, visit Cloud Four’s new PWA Stats.


For more articles on mobile web performance see:

How video impacts mobile web performance and UX, part 1: data and download speed
How video impacts mobile web performance and UX, part 2: autoplay and audio
How to fix your bloated mobile website: fewer, better, smaller images
Optimizing images for mobile: right format, right size, right place, right device
How JavaScript impacts how fast your page loads on a mobile device

Andy Favell is Search Engine Watch’s columnist on mobile. He is a London-based freelance mobile/digital consultant, journalist and web editor. Contact him via LinkedIn or Twitter at Andy_Favell.