CRAP-alla-paZOOPLA – CRAP Talks #10
Now into double digit CRAPs and counting, we thought we’d pull out all the stops and treat everyone to double the amount of speakers from the hosting company as well as double the amount of non-industry related talks too! This edition of CRAP was held in the ZPG office, which houses the all brands within the group – Zoopla, uSwitch, Money.co.uk, PrimeLocation and Hometrack. Tonight, we heard all about conversion rate optimization, analytics and product from the first two.
Ironically, the location wasn’t the easiest for me to find, however, when I arrived, the viewing experience was worth it. With beautiful plants and flowers literally lining the walls of the entrance, it was amusing to see blue plaques of employees decorating the walls as you ascended the stairs. Not only did it showcase the talent they’d retained for so many years, but also was an insight into how prestigiously the company values their employees.
What seemed like a small space before the attendees arrived, there was certainly enough room for the Crapsters. During the introduction, Bhav started by introducing the crowd to Sam Dufton, a Senior Data Scientist from Crisis, the homelessness charity and as we would discover, a speaker at the event too.
He explained that as part of the night it was our aim to raise £400 for the charity, particularly as it was so close to Christmas. In an act of setting an example, Bhav was the first to publically make a donation on his debit card, a behavior which wildly caught on – the second being made before he even left the stage.
Our first speaker was Liam Taylor, a Senior Data Scientist, who was addressing the topic of the importance of communication during an A/B test. He proposed the advantages of two methods used at uSwitch – ‘plan to peek’ and sequential analysis.
The plus points for the former were more to do with quickly being able to see which variant is performing better and having the ability to stop the test earlier. The benefits of the latter included allowing for quicker decisions to be made, understanding the graphs easier and producing more statistically sound results. A seemingly unpredictable topic for the job title, but all the same demonstrating how those who specialize in quantitative data can add value to those in qualitative data.
Well positioned to follow this next was Madeline Wright and Joe Frazer speaking about advanced UX methods. They spoke together about their favourite research methods, their successful experience with using diary studies in a unique way and how they used qualitative data to drive decisions, which they argued provided the WHY and WHERE to dig with quantitative analysis.
To understand more about their users motivations and attitudes toward the early discovery of a new product, Madeline explained that they used face to face as well as telephone interviews to guage the success of a hypothetical experience. Noting the possible issues with this, they spoke about the memory being fallible, and actions on the website by the customers being clearly contradictory to what was said. They also spoke about a piece of software called What Users Do which captures a live end-to-end flow of a journey, which also allows them to measure MVPs and competitor analysis early on.
As these research interviews take place in their on-site lab, which includes a detective-style one way mirror where they can watch their users’ prototype sessions, they mentioned how they have to be mindful of making the data biased in comparison to a remotely based test. After all, customers are rewarded for their honesty in these types of scenarios, but how could you be if you were inside the host’s premises? Surely your honesty would be led by your environment…?
Then Joe revealed a game changer — something to solve the issue. A digitally recorded diary of the user’s experience through Whatsapp. Any activity the customer does in their search for finding a new property is recorded live in the moment through this messaging service. This not only creates less opportunity for the memories to degrade but gives the ZPG team examples which were richer in detail.
They concluded by saying that in order to design better products and optimize the journey for all, these research methods have given ZPG a deeper understanding of the differences between their target groups (treating renters differently to buyers of property) and helped steer the strategy to a successful outcome by applying a not one size fits all approach to their UX.
A short and sweet talk about engagement and retention frameworks came next by Veronica Hulea, Head of Analytics for the property portal team at ZPG. She began by taking us through a traditional framework for this approach, which consisted of acquisition, activation, retention, revenue and referral.
ZPG’s approach decides to focus more on beginning with the activation, followed by engagement and ending in resurrection – arguing in their experience that retention should be an outcome of all these three stages. This not only shortens the lifecycle of a customer in a positive way but ultimately results in a more valuable journey for them and therefore opinion of the company.
Following the interval where I myself queued up behind the masses to donate to Crisis, we had Claudio Caponera, an Insight Analyst at uSwitch. He spoke about predicting switching share in the energy market, which included a live demo of their internal bespoke data tool, Insight.
Using this daily to optimize performance, he told it that it’s also used externally by energy suppliers in the market to predict the performance of their tariffs, even if they haven’t yet been released. I imagine this is incredibly helpful in a market which fluctuates so much!
Insight’s main model focuses on two elements – competition and brand strength, which makes sense as one compliments the other. We learnt that how energy companies are ranked is based on savings from the customers’ current supplier, and not on individual lowest price, given the nature of a comparison site (!). The appearance rate of a supplier on the platform is dependent on the customers’ consumption of energy as well as on location within the UK.
The main learning to be taken from the regression lines he presented us with was a simple one – tomorrow = yesterday + a new modified tariff. An effective formula from a very complex tool!
Next up was something far less commercially minded, moving into the ‘something different’ speakers section of the event. Sam Dufton, a Senior Data Insight Analyst from the national homelessness charity Crisis introduced the NPO in a valiant and proud way, focusing on the crucial goal of ending homelessness together, an aim the audience had been contributing to financially all evening.
He was speaking about the challenges of having limited data touchpoints and information about your customers when working for a charity in comparison to a profitable organization. As much as he mentioned the lack of regular donators was not a concern, it seemed upsetting to know that although the data they’d gathered showed 55% of people in the UK were open to donating to homelessness, only 1.4% of the population were on their database.
An enduring project for them has not been on the quality of donations coming through, but how their supporters interact with them, trying to mirror their behavior to increase their levels of engagement and gain more supporters. An interactive exercise we did as a group involved everybody standing up and we would sit down in their seats as Sam listed all the ways people would’ve responded to their charity through different advertising methods. It also left those standing who were more likely to donate outside of their peak Christmas time and more year-round, without being influenced to donate beyond the more traditional channels he had mentioned.
In addition to fundraisers and campaigners, Sam expressed it was also important to employ more intelligent data analysts to help hack ways to this goal. Using geolocation data (playfully calling it Cri-GIS) to understand where best to build their retail charity shops, using analysis to remove prospects from their advertising campaigns risk-free and make dashboards accessible to all in the organisation (Everybody In initiative) are evidently smart approaches to the lack of big data.
Lastly, but not least, was another non-ZPG related talk by Jess Morley, a social scientist (with data on top of it!) from the Department of Health and Social Care (or, which I preferred, “the Boss of the NHS”). She introduced herself by saying she had met Bhav at a (as Jess put it) “Gender in Tech” talk which I was lucky enough to have also been in attendance at a few weeks prior. It was a delight to see her take the stage at CRAP and talk further about the social ramifications of datafying the NHS, brilliantly titled “The Algorithm Will See You Now, but Does It Care If You’re a Dog?”
She began with highlighting the dangers of suddenly revolutionizing a model which has been in place in the national health service for seventy years, especially when it relies on doctors using symptom checkers online for patients rather than using their medical training.
She also spoke about the fascinating hypocrisy behind the Hippocratic Oath (a protective ‘do no harm’ pledge and the philosophy on which all medicine should be based) whilst women are still regularly misdiagnosed for serious health conditions based on historical stereotypes (i.e. hysteria and wandering womb).
She also shared her horrifically painful experience with endometriosis to bring her case study to life – a medical issue which affects 1 in 10 women, takes on average 7.5 years’ to diagnose and causes the same loss to the economy as diabetes yet is incredibly complicated to detect without an operation. She described her own troubles with the chronic illness as well as some of the outrageous excuses she’s received from doctors in her own journey for closure and relief.
She questioned whether algorithms were truly objective and whether handing over responsibility to data science was a solution to the problems in the NHS. Does all seeing equal all powerful, and is it worth the risk to have these operational efficiencies? To evidence this visually, on a slide she showed the New Yorker’s most repeated and reprinted cartoon in the history of the magazine since its release in 1993 showing a dog on a computer with the caption “On the Internet, nobody knows you’re a dog”. That may have been relevant in the nineties, but now, can you be completely anonymous online? The reality is this utopia does not exist.
We can’t deny stigmas exist and those who are inventing these AI solutions are all co-founded by white men (Babylon Health, Medopad, Deep Mind etc…), another alarmingly strong visual representation of this on its own slide. Whilst Jess posed they may not all be aware of the bias, society is in the loop and we all need to be responsible for questioning issues that exist in the current system.
Her ending comments were powerful ones – we are all being shaped by data whether we like it or not, and we need to be thinking about six key points all the time in the input of our data to society to ensure we gain the outputs we all want.
These were being transparent, being ethical, being accountable, being explicable, being regulated and being fair. Although GDPR seemed to have solved a lot of legacy data problems for other industries, this doesn’t seem to be the case for the health industry.
All in all, we all thought ahead of time that this CRAP was to be a busy bumper one full of opportunity for mishaps, but as usually it went off smoothly without a hitch! Speakers abided by their time slots and the diversity of content provided a thought-provoking evening full of takeaways both commercial and personal. Not only that — we raised a total of £320 for Crisis, but remember, giving to homelessness should be for life, not for Christmas…don’t forget to donate here.