TMW #012 | Behind Spotify's wrapped experience, Optimizely going server-side at Opticon and e-commerce browsing models
Welcome to The Martech Weekly, where every week I review some of the most interesting ideas, research, and latest news. I try to look to where the industry is going and make sense of it all.
π A Quick Heads Up
There are two versions of TMW, the full version for subscribers and a 50% version that's for everyone. Sign up to get the top #3 stories, chart of the week and entire "everything else" list delivered every Friday for this and every TMW π
πTop #3
π΅ How Spotify built their personalised "wrapped" experience. Spotify this week has lifted the lid on their mass personalisation program for their yearly "wrapped" end of year campaign. There are some interesting takeaways from how they approach a very large scale version of personalisation (they process a decade's worth of data for about 240 million users) by using music preferences and listening history. One of those takeaways is how they plan the project. Spotify are a bit famous for their agile operations model but it's interesting that for "wrapped" Spotify work more in a waterfall fashion, mapping out in reverse all of the steps to create the code, build the data models, release to the app stores and collect data. What's more when you think about the personalised cards implementation there are unexpected problems that jump out like creating a version that can handle multiple languages, or if a user only listens to podcasts. All of these are edge cases but for Spotify every scenario is taken into consideration.
When it comes to large-scale initiatives like these the ability of the team to pivot into alternative solutions is key. One such issue was that problem of rendering personalised images in real-time, to address this the team did a whole bunch of planning up front and they were quickly able to identify that if they relied only on back end database systems to version and render the images it would be too costly and at risk of failing on their real time promise. So their development team spun up a proof of concept application to render content as the data is pushed to it to allow for flexibility and close to real time rendering for the user. For Spotify, initiatives like these are a source of learning, innovation, and constant improvement for their development teams which goes to show that sometimes a whole bunch of upfront planning can yield some great benefits. Link
π§ͺGoing all in on server side testing at Opticon. The annual Optimizely conference was on this week with a global and European version. There are some interesting developments in light of the recent acquisition by Episerver. Yet no doubt Optimizely are continuing to double down on product development and server side engineering with Full Stack which was launched in 2016 and has now become their core product set for companies with increasingly mature experimentation needs. One of the best upgrades is the ability for developers to monitor their features flags in real time, to ensure code is behaving correctly but also to get a feed of analytics to detect the customer impact of the flag. Moreover, Optimizely has made some very cool updates to their data platform product with the new stats engine service which exposes Optimizely's AB testing statistical model via an API.
This means you can run the model in other non-Optimizely data sets like analytics platforms along with better integration into cloud products like snowflake and AWS, for things like running Amazon Predict machine learning models in an Optimizely environment. Without a doubt Optimizely are positioning themselves as the champion of server side experimentation and the experience point solution for personalisation. They are going vertical, integrating like crazy and anticipating that digital teams will want all of this in the near term. Link
π« Next in e-commerce, browsing market places. This one is in the subscriber version of TMW. Sign up here to get a link to the full version.
πChart Of The Week
This week I look at how Trump and Biden are spending on Facebook ads, and what it means for political advertising. Available for subscribers of TMW. Sign up here to get a link to the full version.
π Everything Else
Emarsys and SAP. Emarsys, the popular e-commerce marketing automation and CDP platform has been acquired by SAP. This just shows how important customer experience is to SAP. Link
Google's regulation woes. A deep dive into Google's regulation problems. Tech companies now collectively spend more on lobbyists and lawyers than they do R&D. Link
Ikea and CGI. Ikea's product catalogue and now their influencers are all generated by CGI. More companies are doing this, because content can be produced anywhere at lower cost and with near realistic effects. Link
CDP + ML. There's a trend going where customer data platforms are adding ML feature sets. The reason is that CDPs have the capability to enrich and clean data like no other type of platform. Link
Instagram Messenger. Facebook is creating the moat they promised back in 2019. Instagram and messenger are now going to be integrated together. Link
Advertising and hate speech. Advertisers are now stepping in to define for themselves what hate speech is when it comes to social media advertising. Link
Netflix and the news. Comparisons between the rise of Netflix and the fall of traditional news media. It's mostly about operations. Link
Attention economy. A new book analysing how technology companies financialise consumer attention and how it's heading towards the direction similar to the GFC of 2008. Link
The future of Substack. I did a rant about this one on twitter. An analysis into how email newsletters as a media movement has exploded and how Substack is not prepared to capitalise on and maintain it. Link
Baby Bayesian. A children's book on Bayesian statistics. TAKE MY MONEY. Link
Stay Curious,
Make sense of marketing technology.
Sign up now to get the full version of TMW delivered to your inbox every Friday plus an invite to the slack community.
Want to share something interesting or be featured in The Martech Weekly? Drop me a line at juan@themartechweekly.com.