乌鸦传媒

Great basketball teams and great personalization teams have a lot in common.

Imagine a shopping experience that’s completely generic. Worse than generic, it goes out of its way to recommend things you don’t want. It recommends actions that are the opposite of what you’re looking to do. It’s perfectly frustrating. How long will a business based on that sort of experience last?

Now imagine a personalization experience that knows you so well it’s constantly providing you with serendipitously delightful experiences. You’re discovering things you never knew you wanted. But you’re never allowed to use it because the experience never sees the light of day. The MVP never becomes an available product.

Both scenarios are terrible. Unfortunately, a variation of the second is more common. 77% of AI and analytics projects struggle to gain adoption. Fewer than 10 percent of analytics and AI projects make an impact financially because 87 percent of these . What if we could flip the odds? What if rather than most recommendation projects failing, most of them succeeded? Cross-functional, product-centric, teams can do just that. It鈥檚 how innovators like Amazon and Netflix were able to succeed so quickly and so often in their personalization programs. It鈥檚 also been critical for the dozens of successful personalization programs we鈥檝e delivered at 乌鸦传媒.

Recommendation experiences

Everything is a recommendation. That insight came from Netflix: 鈥渢he Starbucks secret is a smile when you get your latte, ours is that the website adapts to the individual鈥檚 taste,鈥� said Reed Hastings, co-founder of Netflix. Recommendations weren鈥檛 features or algorithms. They were the experience; the means to delight, surprise, frustrate, or anger customers. At Amazon, Jeff Bezos鈥� original goal was a store for every customer. This wasn鈥檛 AI for the sake of AI. Both companies made personalization central to their experiences, and personalization enabled Amazon and Netflix鈥檚 visions for more innovative, delightful, and serendipitous experiences. Recommendation experiences (RX) were critical to customer experiences (CX). Experiences were the product. Building products is hard. Josh Peterson co-founded the P13N (personalization) team at Amazon. He described the early days of Amazon as challenging because the company was siloed. Design, editorial, and software engineering were fragmented. 鈥淚t was really hard to ever get anything all the way out to the site without begging and borrowing people from silos. The one time it was always different was when we did a product launch鈥� So, if there was a big enough effort like launching music or auctions then you had permission to borrow everyone to put together your team.鈥� In the early days of Amazon, there were many engineering efforts around personalization. Even though these efforts were led by brilliant engineers, they saw limited success. It wasn鈥檛 until after the launch of Amazon Auctions that personalization made a real impact.

After Amazon Auctions, Peterson and Greg Linden looked to make Bezos鈥� vision for a personalized store for every customer a reality. The goal was a team that could 鈥渙wn its whole space,鈥� to break silos to create a cross-functional team to rapidly experiment and deliver. This was the first team, outside of the design organization, to have designers in their team embedded with web developers and technical project managers. This enabled a higher number of launches compared to other teams. The impact of their model was so successful that it became the basis of Amazon鈥檚 famous 鈥淭wo Pizza Team鈥� approach 鈥� essentially a team small enough that they could be fed with two pizzas. Small teams that were decentralized, autonomous, and were 鈥渙wners鈥� of the business could move faster and launch more experiments. More experiments would enable them to have more successful innovations.

Experimentation

Successful personalization teams foster a culture of experimentation. Creating a culture of experimentation requires diverse, multi-disciplinary teams. Below we show the various skillsets and domains that are required for modern personalization teams. The circles don鈥檛 represent people, they represent skills. Great basketball teams and great personalization teams have a lot in common. In basketball, you need defense. You need offense, both close to the rim and from afar. You need diversity in skillsets. You could get lucky and find a unicorn but fielding multiple teams of unicorns is not practical. Creating a team of all-stars sounds good on paper, but there are plenty of examples where those super teams fail to live up to expectations. A team without a diverse set of skills is unlikely to be very successful, and almost certainly not great.

鈥淓xperimentation requires blending creativity and data. Practically, this becomes a blend of statistics, behavioral economics, psychology, marketing, and expertise in experience design.鈥�

Small teams with most of the skills above are more likely to do end-to-end personalization well. No one person will have all the skills needed, but together they鈥檒l bring more experiments to the table. Early Amazon teams were engineering and data-science heavy. It wasn鈥檛 until the addition of design, business expertise, and a product-centric approach that they were able to execute end-to-end and achieve Bezos鈥� vision.

Velocity is a leading indicator. Successful personalization teams test many ideas. They break experiments into small chunks so no one failure is large enough to disrupt the business. They test and learn quickly. Testing a dozen ideas and refining them will be more efficient than trying to make one idea 鈥減erfect.鈥� Our intuition on what is going to work is often wrong. Testing many ideas allows the data and results to guide us, rather than intuition. This requires personalization teams to develop many ideas end-to-end quickly.

What does the future hold? Cross-functional, product-centric teams are the beginning, not the end. Experimentation requires blending creativity and data. Practically, this becomes a blend of statistics, behavioral economics, psychology, marketing, and expertise in experience design.

These teams need to track which features drive results to understand what is working and what is not. The goal is to achieve consistent and reliable serendipity from personalization efforts. The obvious is not serendipitous. Experimentation is needed to discover that which is not obvious and that which drives business outcomes. Without that, we can鈥檛 scale serendipity.

INNOVATION TAKEAWAYS

DIVERSITY LEADS TO SPEED

Speed leads to innovation. Diversity leads to innovation. End-to-end cross-functional teams with dedicated resources are more likely to successfully implement personalization programs and innovate faster than their peers

A CULTURE OF EXPERIMENTATION IS CRITICAL

Velocity, variety, and volume of experiments are leading indicators of innovation. 鈥淥ur success at Amazon is a function of how many experiments we do per year, per month, per week, per day.鈥� 鈥� Jeff Bezos

SPEED IS A COMPETITIVE ADVANTAGE

Testing and learning iteratively as well as being able to deploy quickly contribute to faster speed to market. 鈥淐ompanies rarely die from moving too fast, they frequently die from moving too slowly.鈥� 鈥� Reed Hastings

Interesting read?

乌鸦传媒鈥檚 Innovation publication, Data-powered Innovation Review | Wave 5 features 19 such articles crafted by leading 乌鸦传媒 and partner experts, about looking beyond the usual surroundings and be inspired by new ways to elevate data & AI. Explore the articles on serendipity, data like poker, circular economy, or data mesh. In addition, several articles are in collaboration with key technology partners such , , and . Find all previous Waves here.