Monday, 16 April 2018

Further developing an onboarding process for a green field product

This is part of a series about my side project Bashfully, which aims to give graduates and other new entrants to careers a seasoned professional level way of expressing themselves through the super power of story telling. Following the core principles of being discoverable, personalised and guiding in approach.


Photo by Etienne Boulanger on Unsplash
Following on from my post on Building an onboarding process for a green field product we have building the experience. One of the lessons I pulled out previously was about launching something to get feedback. Even if you don't feel ready. It's easy to know the theory, but hard to put yourself out there!

I'm really glad that we did as it allowed some feedback and integration issues to be tested while we polished.


Background 


One approach that we have taken is to slowly refactor the experience as we add functionality into the edit screens. To start with we had a limited set of data editable. Basically the story elements. The "extras" like social links we didn't include in the first release. Now there is a unified experience, with a better underlying code base.

An interesting choice that we had to make was the "minimum" amount of detail people had to enter before "completing" a profile in the setup. In the end we decide to go with almost nothing. Make the sign up as easy as possible. 


First iteration

In the first iteration we provide no onscreen guidance and relied on the user going to "Manage profile" from the action menu. Then finding what hadn't been completed. This got me interested in empty states.

I have some experience of this with data driven applications. These are mainly the sole focus of the screen and I have tackled improving them by providing context, for example on a log out screen not only providing a log in link but directions to other brand touch points. The experience here with Bashfully has given me a bit of twist to the way I look at them. The data that was missing has three different uses, with different impacts to user goals when they are not completed.

Looking at how similar products do this, YouGov and LinkedIn give you a percentage completion figure. This can be useful in seeing how much more that you need to complete, but it doesn't give you any information on the impact to assess what to do next (if you lack time or motivation to do it all)

Second iteration

So next, we give feedback on what hasn't been completed and the impact. So, for example not choosing a profile URL means it isn't discoverable. We also choose not to spread the state messages over the profile screen, where the content should be, but to position it to the profile owner at the top of the profile text.

We've now pushed this live so that we can monitor what our users think is most important. Starting with the top three features of the site. There are other features where we can add a status indicator, for example if any skills have been tagged.


Future improvements

Another avenue we need to explore is doing something a bit more helpful with data imports. At the moment we populate the core fields in experience, but there are some ideas on how we can infer some helpful things to recommend to new uses completing their profile. Oh and extending the "LinkedIn profile import" with other integrations like Facebook, Codepen, and Dribbble.

Overall it feels like the approach of batches of experimentation at the same time that we improve the code base is the way to go. Otherwise, it's too easy to rush experiments with "temporary code" that becomes technical debt you need to worry about. The thoughtful way that Martyn is evolving the architecture deserves a lot of credit for enabling this. 

One thing we really need to watch out for though, is not gold plating the onboarding process. We need to keep an eye that experience matches the same level of functionality once completed. One way to do that is to develop another mini feature experiment while collecting data on this one.

Further Reading

Thursday, 12 April 2018

WEBINAR: Five Strategies for Getting the Most From AI hosted by MIT Sloan Review

Super short post to tell you about today's MIT Sloan Review hosted webinar "Five Strategies for Getting the Most From AI " based on the blog post of the same name by Jacques Bughin. This shared the results of various surveys and research by McKinsey & Co.

From this research, the highlights on developing a successful AI strategy are:

  • you need to be digital based and native already, no leapfrogging straight from an analogy business
  • don't try and do it on your own - they are not mature technologies yet, so use an ecosystem with startup partners and academia
  • be bold - take the chance to reinvent your products

The other theme running throughout was that when developing strategy you should think about growth as this lead to bigger profit gains than cost cutting. The advice was also to do it soon, from the companies survey there was a bigger boost from early adopters who innovated.

Finally, it's about humans as much as technology. The top three reasons for why companies may not take up adaptation were about capability to use the technology, for managers to embed it, and lack of the right people skills to complement the technology.

I was pleased to see that theme being so prominent. I feel that "Augmented Intelligence" has an important role to play. It can be a powerful tool to help us do our jobs better. Let's get computers to do the things that they are good at - pattern recognition for one - and allow us to concentrate on what we are - moral reasoning and creativity for starters!


Sunday, 11 March 2018

When SEO meets the MVP process on Bashfully

This is part of a series about my side project Bashfully, which aims to give graduates and other new entrants to careers a seasoned professional level way of expressing themselves through the super power of story telling. Following the core principles of being discoverable, personalised and guiding in approach.

So after getting the SEO infrastructure sorted out we are back into an experiment and observe phase. 


LinkedIn on Google
 LinkedIn is the yard stick that we need to beat. The features that we are honing in on based on the MVP process are discoverability, customisation, and guidance. These allow us to add value as a David fighting the awesome network effect of the LinkedIn's Goliath. As an example to the left is a search result for me going to my LinkedIn profile. There are a couple of points that I like - my name, job title, location, and current employer are all easy to read. The thing that I don't like is that the description is very impersonal. Is the fact that I have had 7 roles really the most interesting thing about me? Or is it doing more to tell LinkedIn's story? Is this their SEO not mine?

I'm not just picking on LinkedIn here, it is the best out of the other profiles that I have. AngelList, GitHub, and Flickr are all very fact-and-figures based. 


Bashfully on Google
Now a look at the Bashfully results. Taking the LinkedIn good points, this needs the current role. But I think that we have done a good job in the content organisation and meta data to put the story front and center. This puts something of my story and aspirations right there. With my current role and employer I think it will round it off nicely.

One of the problems in testing this is that Google is very much a black box. You can't tell them how to display things, you have to give them hints and hope that they take them. It's actually very much like being a Product Manager, working with influence and not direct authority.

The main thing that we started testing this week was the rich data structure to help pull together a profile owner's digital footprint. Unfortunately, the Google Search Console doesn't report on the rich data for people or organisations. There is a "jobs" type but it has required fields that we don't, or want, the data for.

The lesson to learn from this is to get the basics right. Structure your pages so that Google picks out important content. Then use CSS to visually represent this to your uses. Next you can build on hinting to Google and crafting search result snippets.


Edit2: Second lesson was that the Google Structured Data Testing Tool doesn't flag all the errors that the Google Search Console does. Somewhere in some refactoring of the solution, as we learned from what we needed and the results started appear, the "name" got lost. This was quite an important bit of metadata to not be there! So, even if you think that you have got it right, test the live site and watch the results separately. 

Other search engines

So far we have been very Google focused, and had some success. Now it's time to move onto other engines. We are starting with Bing and DuckDuckGo (which uses Bing as part of its ranking/indexing system). Bing actually has quite a nicer result for LinkedIn than Google. They support a similar range of markup that we already have on the site, including Open Graph and Schema.org.



Bing example search result

On the other hand DuckDuckGo has security as its main feature, so the results are rather sparse. It is a nice touch to put the sites logo alongside the URL though.
DuckDuckGo example search result
Edit1: As a minor success after registering with Bing webmaster tools, I have managed to get Bashfully to appear on the first page for some key search terms! This is mainly seems to be down to the OpenGraph data and structure/content being relevant ... but not the individual profile pages yet


Further reading


Sunday, 4 March 2018

What I've been reading w/c 26/02/2018 Innovation and Product Culture

Great look here at Product Analytics. Think I use about 5 tools altogether, and even with Google Analytics, I layer other tools on top to help make the data usable Life Beyond Google Analytics: Pick the Best Tools for the Job

Thinking about product culture started with the start of a new series on Medium from FutureLearn - Using agile principles to develop company culture Part 1: Introduction and it promises to be a great look at how a successful organisation in delivery can live the values of the agile manifesto. It was then a short step to Stop Blaming, Start Innovating a great article that teaches us that Innovation, like charity, begins at home. Thoughtworks have a similar take and say that 

"Innovation is the key to unlocking a best practice culture" Thoughtworks, 2017

Next up were two posts that cover more of the nuts and bolts of Product Management work. The first was a round of top tips on How to become a great Product Manager, according to 16 of them. The next was  The Problem With Being a Fix-It Product Manager really resonated with me. Sometimes you just need to treat your career like a product, make sure its purpose is clearly defined and focused. Not just for your own sake, but to ensure that the product culture around you works and isn't just making up for organisational gaps.


Thorbjørn Sigberg writes in A funny story about projects

The only funny thing about projects, is that they tend to go horribly wrong in ways nobody anticipated.
and weaves an anecdote about large scale IT projects with infrastructure projects as he sits on a delayed train. I've certainly been on a few projects that match that! (as well as a few were it was obvious at the outset). The lesson I take from this is to keep outcomes and decision making as closely together as possible, these allows change to the things that nobody anticipates. So that's projects ... what about innovationWhy do Innovations fail? Not as big a takeaway here, but similar theme in too much bureaucracy and getting to the job to be done, to help the maximum number of people win with your product.

Can’t believe I missed National Storytelling Week earlier in the year! Interesting findings in Britain's Biggest Companies Fail to Tell Their Stories Clearly that neither CEO reports or “About Us” pages have much emotional language. Missing a trick to form a connection on a human level. When you consider Stephen Shapiro's advice


they are missing out on a massive opportunity to be part of what drives buying behaviour. So, are missing out on getting to the job to be done and being part of the users journey.

And finally, this week saw the start of a new section on the 15below blog the "15below Tech Take". Here is a piece where I look at Blockchain and #TravelTech. There is much more to it than currencies! so much going on in this space at the moment, the next big thing for passenger communications?

Sunday, 25 February 2018

Venturing into the world of SEO

It has been a bit of a voyage into the unexpected in looking at SEO this month. For my side project Bashfully, which is to create an online profile for people early in their careers that has three guiding principles. That it should be:



  1. Discoverable - people need to be able to find the person based on their skills, experience, and aspirations.
  2. Personalised - the skills and experience need to have the ability to be tailored for specific job applications.
  3. Guiding - given the above, give enough structure that allows the profile builder to tell their story in the best way possible. Also a longer term goal here is to provide feedback based on other profiles that match their aspirations.
The features that we develop tend to rotate around each of these goals to keep the product balanced. We hadn't done much in the discoverable area, apart from setting meta data required for creating the cards used in sharing to Facebook or Twitter. Since this came up in our user research we needed to start improving. So, that's the why, now on to the how.


Google Search Console

The first step was looking at the search terms used to reach the site in Google Analytics. Looking for custom reports to help dig into the data, I learned about Google Search Console. This required some extra setup, with linking to Analytics being enough to "prove ownership". The Google Search Console appears to going through an update at the moment. The new design doesn't yet cover all the functionality.

Looking at the data there it was a bit bare. Not many pages were indexed, so therefore not showing up in results. Under the "Crawl" menu was a setting for sitemaps. I knew that Bashfully did not have a sitemap yet, so that was my first stop. 


Sitemap

I wasn't entirely sure what format the sitemap needed to be in. So I went to free site map generator.com. This is not the most beautiful site that I have ever encountered. But it did generate an XML sitemap for me. A little bit of tidying up I put it in the root of the site. One extra little step was to add the path to the sitemap in robots.txt




Digging into results

Once the new sitemap was up and running I went to "Crawl" > "Sitemaps" to get an index started. Within a couple of days we started seeing the clicks that were showing up as hits in Google Analytics. However the whole experience isn't as obvious as say Google Analytics. Which is why SEO tools that take the Google Search Console data exist. 

SanityCheck.io is one such tool. Once you give their user permission to access your search data then you get three options to guide you:

  1. ...get some new content ideas
  2. ...find striking distance keywords
  3. ...improve the CTR of a page




I found this really helpful. Especially the "striking distance keywords", this helps find words that are on the second page of results. I only got the chance to put one small text change through and view the results before the free trial ran out, but I did see more varied search terms match with the site.

Our main approach here has been to make small changes where we can measure the results and probe to learn more. We don't have any SEO experience between us, so doing large changes would not be a great use of time ... and just as likely to make things worse.


Other considerations

Speed that pages get served up is now part of the Google ranking. This is alongside the other changes to nudge sites to be mobile friendly. SanityCheck also allows you to setup speed checks on specific pages and caught a page slowing down in the desktop version - but strangely not in the mobile test. It is strange to think that websites wouldn't be responsive these days.

Appearance of the results. This was a bit more of a rabbit hole! Rather naively I thought that  Google might use the Open Graph meta data that Twitter and Facebook use for enriching links. But no, they use Structured Data. This is really useful in emails for events etc and as it turns out giving hints on how to display search results.



As you can see from the example above it does duplicate the data already there, but it does have some key features that we are interested in. One of the top requested problems to solve in our user research was being optimised for SEO. These structured data and the "Person" type gives us the best way to give Google the hints it needs to put forward our users stories in the best possible way.

One of the frustrating aspects of structured data is that is only supports a subset of the full spec. So there was a bit of trial and error to see which other properties Google supported above the few mentioned explicitly in their intro site.

Further reading

Sunday, 11 February 2018

We need to talk about Alexa: Common use devices in a personal world

Photo by Andres Urena on Unsplash
This week I'm going to reflect on a year and bit of using two voice assistants - Alexa and Siri. Although much the same would apply to Google Home. I must start by saying I love Alexa and the echo dot. She does just enough and is unobtrusive enough in my life that I'm not a slave to her ... in the same way as a smartphone.

Last month I wrote about one aspect that "we" have not looked into enough - privacy. 


“I think it will make for a perfect alarm clock”  Trusted Reviews - Amazon Echo Spot
Here it looks like part of the problem with technology is the uncritical approach of what could go wrong, in building and selling. There is no mention of privacy concerns apart from throw away comment about a "mute" feature. As well as the obvious issue of an internet connect camera in our bedrooms. I have been thinking about other issues related to a mindset used to personal in more common use areas. So I have done some digging into people's opinions of the challenges technologists need to solve.


What voice assistants give us

But before I start looking at that, here are some of the good things these products give us:

  • Timers when we have our hands full, great for cooking!
  • "Routines" on Alexa can be used to create really personal alarm clocks - surprised smartphones haven't had these for years
  • Control over music

This is what we are trading for the areas of concern below. With 17% feeling that nothing needed changing, and a further 8% wanting more access to these assistants on the move! Which could be done to where people are in their stage of life? And what they perceive that they have to lose. One of the qualitative responses that I received suggested this could be a factor:
"I'm at that age were I'm not particularly bothered. In fact, I'm probably boring them rigid with my timers, alarms and reminders." Mark on Ipsos i-Say


Privacy

Two quotes that sum up the responses of the 41% people worried about privacy I got were:
"It’s a government listening device" Ben via Facebook
and 
"... just like all the new 'smart' products like TVs, Meters, fridges etc they all made for spying on us 👎🏽😣" veganpanda on ipsos i-Say

How else is privacy affected in these areas? Well, since Alexa has added messaging anyone (including 6 year old nephews!) can get access to messages meant for specific people. This weekend I have noticed this when Alexa couldn't understand a command. She fell back onto checking the messages.

This is wild speculation, but I imagine the main use case used in tested against is adults. Either in a close relationship living together or single people home alone. 

The other thing to think about is that you don't always know Alexa is there - is she in that speaker? Or maybe that lamp? We are moving towards a connect world. Are you happy with the privacy considerations as they are?

Security

To sum up the 34% of respondents who voted for security:
"There are too many risks the devices can be hacked" - Marmum63 on ipsos i-Say

This is very like the above two cases. Most people I know disabled the voice ordering shopping option as soon as the devices are setup. Since the first thing most people do the first time they visit someone with an Echo is to order something bizarre.

The other aspect is that you are giving one, maybe two or more corporations access to you home. Constantly listening (and maybe watching). With very sensitive microphones connected to the internet - what could go wrong?

Lots of people - me included - find this technology really cool. And the companies that make them are in relatively liberal democracies where the rule of law is valued. But what if that isn't the case in the future? What about these devices being used in countries that already look to control the population with technology?


The family experience

One of the interesting quotes on its suitability came from Facebook:
"I wrote a skill called School Run to help make sure everything gets done. Was planning on entering a competition. But we tried it and it was a hideous experience for the whole family. I think voice tech has a place but maybe not in the home. " Neil (another one!) on Facebook
and another on the downside of giving digital assistants human personas
"We changed ours to respond to 'computer' because I didn't like the way it made it fell like having a servant called Alexa." Helen on Facebook

Another issue with the common use aspect, which I've experienced, are recommendations. They work brilliantly when you log into a website on a device that you control to use the way that you want. But if I use my wife's Echo connected to her Spotify to play songs that only I like. Then it starts to devalue those recommendations to her. Likewise with YouTube on the Firestick. I use it to watch music videos or travel industry related conference feeds. Yet because one of my nephews likes Peppa Pig I get many more recommendations based on that!


Conclusion

I remain cautiously optimistic about the promise of smart assistants and breaking free of screens and keyboards. But it is still very early days. It feels like we are in the era of PC voice dictation software. Or the early internet enabled phones that used Java apps and WAP pages. That's to say the current crop of voice assistants, which don't need training to understand you, and capabilities of native apps on smartphone now are an order of magnitude better. 

What it took to get here was to create new ways of thinking and working with products suited to these devices. Not to copy from the desktop paradigms and local processing. The voice processing for Alexa benefits from having vast amounts of data in the cloud. alongside a feedback mechanism from the users that makes my copy of Dragon Dictate in 2000 look like a toy.


We are still very much find our way as an industry in the privacy and ethical challenges that common use devices over personal ones provide. We need another leap in technology to detect who not only what is being said. If we solve these then we will have a brilliant benefit for easing the friction of using technology. For example, checking flight information in a natural way. 



Further Reading

A one question poll hosted on ipsos i-say. This surveyed UK adults in as close a random sample as I could get. I chose this over a Twitter poll due that more varied sample as my followers have a heavy industry bias - UX people, Product Managers, and developers. At time of writing this received 425 responses. The question posed was:

What worries you about voice assistants? (e.g. Alexa, Google Home, Siri, or Cortana)

With the description:
With everything seemingly getting voice assistants from lamps to speakers, what worries you the most about having all these devices listening and watching us in our homes?

With apologies to HBX's Professor Janice Hammond on the question design! I took a decision for an opinionated question based on my review of mainstream reviews of these devices. I hoped that this would lead to thinking more about the downsides. It had the following options as responses:


  1. Privacy
  2. Security
  3. Nothing, I'd even like to take it on the road!
  4. Nothing Happy with the way things work now
  5. (the ability to add another response)
For more qualitative replies I also posted on Facebook, Google+, and Twitter:

Anybody got some good stories about Alexa or Google home type voice assistants? Things you didn’t expect it to do? Ways children have confused it? Or it has affected recommendations from other services that you use? Worries about privacy?


Sunday, 4 February 2018

Second set of learning from the Bashfully MVP process

I was going to write about SEO and Bashfully, but is usual on software projects other things cropped up. In tandem with my day job I have mainly been thinking about maintaining delivery momentum.One of the downsides of small projects teams is the lack of capacity and time. However, the upside is focus and alignment. We manage this by:


  • Keeping a Small backlog, we're open to opportunity but don't fill a backlog for the sake of it. Each completed feature is usually an avenue to learn and build.
  • Referring to the vision and remembering YAGNI, there have been times that we have rejected ideas as they aren't core to what we are trying to achieve. And others that just aren't right yet (which we immediately discard, no clogging up the backlog!)
  • Talking before starting dev, really simple but not relying on story formats or mock-ups. We talk through what we are trying to achieve and what success looks like.
That's the good things that I have learned in this phase. The bad things come down to prioritisation and what we knew at the time. With hindsight I would have been tempted to tack on-boarding and tailoring profiles differently. But it is a bit of a chicken and egg scenario. You need compelling features to get people to sign up, yet you need users to provide feedback.

After we had profile tailoring (what we are calling "lenses") I set up a limited Twitter campaign. This did increase the engagement and got us some potential users, but from the analytics I could see a drop off during the setup process. One of the factors behind not prioritising this earlier was that user research had shown wanting to do something with LinkedIn data (others were asked for, but LinkedIn was far and away the winner).

Worth bearing in mind that no matter how you ask it, people aren't going to give useful answers to setup questions. You just have to make it as easy as possible to get in. The approach we have taken now is to get the bare minimum and then allow a full profile to be built up at leisure. Also no matter how easy you think you've already made it, go one step further!

Which brings me onto the Tweet that wins this week...



Further reading

Further developing an onboarding process for a green field product

This is  part of a series  about my side project  Bashfully , which aims to give graduates and other new entrants to careers a seasoned prof...