JUST EAT CRAP – CRAP Talks #9

We arrived for this month’s CRAP Talks at the Just Eat office in Farringdon to form part of the biggest audience they’ve ever hosted for an event. Whilst we enjoy the spectacularly lofty views of the fifth floor, overlooking the beauty of old architecture with the modern, across the world, two hours ahead of us, the first ever CRAP Talks Istanbul is taking place in Turkey’s capital, of which we had the pleasure of viewing a live stream.

As if it couldn’t feel any more special a celebration of how far they’ve come, this is the first time CRAP Talks has had pizza as it’s snack of choice, and as they’re Just Eat, you better believe they were connected with the best pizza joints in town. The crowd were engaged in high volume chat, in what seemed to be a place too small to be networking in, which just goes to prove the expanding attendance with which this increasingly popular meet-up has to contend.

Joanne Schwartz, Senior Growth Analytics Manager at Just Eat, and a fiery, alert ball of energy and passion, bursts onto the stage, beginning the evening with a talk about analytics and prioritization in a world of fudgy data. Dealing daily with world marketing and product teams, she proposed three questions to ensure every requirement they took for a dashboard build resulted in value-added. These were 1) ‘Why does it matter?’ 2) ‘How does it contribute to KPI goals? and 3) Should we do this above all else?

This opportunity to assess business requests removes any risk of working with what she describes as ‘fudgy data’ that is more malleable than accurate. Developing assumptions to forecast potential KPI contributors is a method she has adopted from her financial trading background, and bringing this in the tech world is an effective tool for choosing an objective prioritization framework.

Using the ICE method – importance, confidence and ease – helps to guide the analyst to focus on growth through numbers rather than just vanity metrics. Having come from in her words, an ‘aggressive culture’ when she worked in Israel in terms of company goals and investment style, she has said this helps her achieve more direct results from her stakeholders, and a culture Just Eat has embraced as a highly successful method of KPI gathering.

Hemini Dhanani, Senior Product Analyst at Just Eat enlightened us with an in-depth overview of Project Slingshot, which I must say was an incredibly well presented talk from a steady and sure professional. As part of a ‘taking risks’ initiative, Hemini assembled a team of UX researchers, product managers, analysts and engineers on a project which would promise Just Eat staying relevant in their market but also accelerating past their competitors, whom UberEats they cited as a major threat from a visual personalization perspective.

By fixing basic product areas such as their recommendations section and offering new deals through other websites (areas of improvement all revealed through extensive market research with their customers) they began closing the gaps on vulnerability. This experiment not only improved customer journey flow on all devices but also helped them to discover the most productive way to achieve these breakthroughs was through a specifically assembled agile vertical team structure, comprised of a Product Manager, Delivery Manager, Tech Lead, UX Lead and an Analyst.

The results of this project included a significant incremental uplift in order frequency as well as a growth in orders. Despite celebrating the key wins on these continuing improvements, Hemini stated the challenges of moving the needle so much on the modernization of the brand personality, design changes and website content given the restrictions of being a regional business. Perhaps a key consideration for the leadership teams? However, these small explosive experiments seem to be an integral part of Just Eat’s approach now and the learnings they present will only further the brand’s triumphs.

Simon Poole, the Head of Engineering for Data Algorithms at Just Eat revealed the secrets in how they use data science practices (namely machine learning) in product discovery, which with over 70,000 restaurants on Just Eat is arguably one of the most important engines of the company. He briefly introduced us to how these techniques are used to predict delivery times, optimize these deliveries and how voice ordering will impact their contribution to the market.

He promoted the view of algorithm development being a discovery in its own right, rather than simply a delivery. This was proven in his explanation of their evaluation process, which involved a complex journey of analyzing truth datasets, monitoring historical analysis and undergoing strong user and A/B testing processes. Simon also encouraged the idea of making data features a first class citizen in your process, as so much can be learnt from the customer in order to simplify the analysis.

The final point he advocated was to embed data scientists into product teams so they are engaged in a mixed discipline environment. This not only allows for a higher performing team but also a correlation of collective intelligence (basically, the more variety, the more success). By having their data scientists spending the majority of their time rotating in different product teams, Just Eat can boast progressive productivity, higher engagement in workload and ultimate trust amongst colleagues. How can you argue with that?!

The final non-industry speaker and non-Just Eat employee, Daniel Hulme, the CEO of Satalia or ‘The Solve Engine’ as he’s known on Twitter, illuminated his view on AI and the end of the world. I’m really pleased CRAP Talks uploaded the video for this one, as I was so engrossed at the time that I actually stopped taking notes. Squeezing an hour’s long talk into 30 minutes, this charismatic, articulate and highly intelligent individual left the room speechless with his controversial and radical concerns about the future.

Using extensive interactive examples which included subversions of the defined technology stack pyramid and contrasting interpretations of data sets, Daniel spoke about applying predictive power to descriptive analytics in order to dismiss traditional weak AI approaches.

He argued that statistical algorithms are unable to solve optimization problems effectively and getting computers to do activities better than humans was a poor attempt at making the most of artificial intelligence. By hiring operational researchers, Satalia applies goal-directed adaptive behaviour to decision-making, which in turn increases optimization productivity.

With a challenging and super speedy first 20 minutes, nothing could’ve prepared us for the extreme turn in the final 10. Once the mention of the Fermi paradox and technological singularity came about, there was no going back. The idea of our final invention being the biggest existential threat to humanity and happening in the next 50 years terrifying enough, and posed the question of whether it was rational to build a brain more intelligent and faster than ours.

As intense as Daniel’s master plan for a decentralized company formation was, it seemed to be the most convincing solution to the war on talent which companies battle in the world today. With the managerial hierarchy of companies ultimately in a defeated state, relationships within organisations who choose to adopt these conventional structures are deteriorating.

By removing the idea of companies completely and creating a platform where anybody in the world is able to contribute to a project and be remunerated fairly for it, this would allow for more freedom to work on interesting problems and grow our talents in ways we choose, for maximum progression of our species.

Daniel’s 30-year roadmap towards decentralizing the concept of Satalia by removing KPIs, managers and rules seems to be possible. His hostile aversion to competing with those companies who possess not only the access to global infinite capital but also superior accumulation of wealth, talent and power, has clearly given this uprising against the capitalist game. To successfully launch this decentralization would hopefully result in the utopian world he presents in his article set in 2048.

Concluding the talks with a heated Q&A session (thank God this wasn’t released on video!), I couldn’t help but admire Daniel’s staunch, unwavering conduct toward those provocating his liberal beliefs. Not only did he embody so well the CRAP Talks intention of being imaginative and original, but it is these individuals who can transform the world into a better place with their transgressive ideals.
I’m sure I wasn’t the only person who left the venue that night believing that CRAP Talks was more so evolving into a platform for revolutions to occur.