I've spent nearly a decade working in the experience layer of eCommerce. I’ve worked with rules-based systems, and more recently have focused entirely on machine learning applications through Particular Audience.
I’m interested in how information on the web is structured (or not structured) and the signals that can help us define relationships between data, to make our relationship with the internet more symbiotic.
Brands and merchants know us as Particular Audience (particularaudience.com), and our consumer brand is Similar (similarinc.com). We have managed to aggregate every item on the consumer web and are really really good at knowing when to show each one to you.
Contextual commerce is the umbrella term we identify with.
Sick of rules-based approaches to personalization and having segment-based targeting prove perpetually ineffective and unscalable, I started reading up on machine learning applications used by Amazon and other big tech platforms.
I was particularly lucky to get my hands on Anand Rajaraman’s book Mining Massive Datasets where I (after some study) had my perspective flipped on a counterintuitive fact in how to do effective personalization.
Personalization is another word for prediction, how robust your data is plays the most significant role in making accurate predictions.
I learned that user-orientated, and especially segment orientated approaches are ineffective for the following reasons:
We’ve honed our technology leveraging a mix of collaborative filtering, computer vision and natural language processing working with some of the world’s best-known retailers on their search and recommendation systems. Creating a different website for every customer, in a market full of static shopping experiences on monolith eCommerce platforms.
The key questions outstanding are, how can this logic work outside of a single website? Can we create some sort of internet-scale recommendation engine? Wouldn’t that make the internet easier to navigate? Can we automate search based on somebody’s context? To answer these we built ‘Similar’, our first consumer product, which you can add to Chrome now at similarinc.com - check it out, there’s a tonne of exciting applications in the pipeline but it will save you a lot of time and money in the meantime.
I’ve been lucky to find myself a part of the most incredible team, we’re now at 40 people globally and it feels like the best mix of skillsets and diverse perspectives I could have hoped to create. Scaling fast meant some incorrect hires, but everyone at PA today is someone I would happily work for in their respective area and that makes for an environment of high standards and mutual respect.
Probably building someone else’s! Back in 2016 I was considering an MBA but decided to spend the $200k and 2 years building a startup instead, it has worked out to be the right decision so far.
We have several products now and gross merchandise value seems to be the common metric to measure our progress as a business overall. We’re currently on a $61.5m run rate and expect to finish the year at $100m.
There are a lot of metrics that play into that headline measure impacting customers and merchants that use our products such as conversion, basket size, margin, cash:stock ratios, landed cost, operating leverage, money saved etc.
I consider revenue a proxy for how well we are serving our customers, and our revenue will grow around 390% this year. Revenue is also a reflection of our product thinking, and product traction, demonstrating our diversification and experiments in how merchants and shoppers want to interact with one another online.
We’ve recently found our first organic loop, meaning a channel that grows without investment. We’re nurturing it and are excited to see what it can become.
I’m most excited about quantum applications for security and encryption. I am most worried about quantum implications for security and encryption 😅.
Probably to skip the investment banking career and get straight into an early stage, high-growth business and learn