Catherine Neasmith: move fast and break things (but not the infrastructure)

Catherine Neasmith has been a force in data in and around Silicon Valley for the best part of two decades, with recent posts at digital behemoths Netflix and Facebook. She spoke at the DMA’s data hub breakfast.

Neasmith recently set up her own data consultancy QBiz. “I’m only six weeks in,” she explains, and is in London partly to talk about her various data roles, but also to visit clients The Guardian.

She opened with her potted history of data. There was a steady progression in the capacity of data stores in the years up to around 2011, when stores could handle more than 100 TB of data. “This was the peak, with secure performance systems for companies. It was a nice place. Not everything was digital yet.

"Then big data happened,” she said.

“Between 2012 and 2014 there was chaos. Demand for data increased and it was hard to control. Companies studied Hadoop interfaces, with servers getting bigger and bigger.

“Now we have old design technology combined with the new stuff," she says.

Facebook

According to Neasmith, who worked at Facebook as a consultant and project manager, Facebook’s approach to data is, “not about technology, but business need. The goal is impact, not insight.

“But everything is captured. How do you deal with something so big?" she asks.

To achieve impact, you need the insights below that. Neasmith said that in the past, data insights often didn’t change anything. “Facebook turned this on its head,” she says. “They have massive data sets channelled into something you need.”

Culture and personality

According to Neasmith, the huge advantage for companies like Facebook and Netflix is, “The single focus on a single product. It’s a results-driven culture.

“Everyone agrees on the key driving metrics and everyone rows in the same direction. In less unified companies this is much harder,” she says.

While at Facebook, Neasmith recounts a story where two managers came to Mark Zuckerberg because they disagreed on a growth metric. "How is this possible? This started a project to unify the metrics. You need unified metrics and leadership. This will in turn generate impact - find something that is focused on goals or products, not just 'interesting stuff'."

Measure what you value, not value what you measure.

"Analysis produces insights which feeds more analysis," says Neasmith, and she warns of companies moving into 'analysis paralysis'.

"At Facebook, the programme is related to results. If you don’t get results, you don’t get your bonus. The motto there used to be 'Move fast and break things', but this has now been upgraded to, 'Move fast and break things (but not the infrastructure)."

It's important to integrate data teams with the rest of the business, she says. "In the old world, data analysts were separate. Embedded technologists can make a huge difference. Lots of data teams are begging for attention and if you are not integrated with the business, you are spinning in analysis land."

Netflix

Neasmith worked for close to three years at Netflix. "There are three areas data is critical: subscription acquisition and retention; content purchasing; and recommendations"

She says Netflix is unusual in that they encourage the churn of customers, but do have 'huge' winback campaigns.

More interestingly, "They use data to drive what movies they buy, and link spend on content to retention. This is not an obvious link.

"We looked at engagement levels, and used a predictive algorithm to help negotiate prices for the movies we wanted. In the end there may be a few movies they want in a tranche, but with some TV stuff too.

Analysis of viewer habits compared 'quality' with 'non-quality' viewing hours. They found a relationship between usage, renewal rates and lifetime value.

"We look at the value of a view - the completeness of watching, how many new vs old catalogue views, etc. This all feeds into a CRM programme, and we could launch campaigns based on usage.

Data and original content

"This is my favourite subject," she says. To the surprise of some, the data team were asked to recommend, "who should play Kevin Spacey’s wife in House of Cards. This was when the data skewed to DVD. We tried to look at outliers, and Robin Wright was on the list we produced. I'm not saying our recommendation secured her, but we were asked," she says.

Neasmith was also asked to monitor consumption of content. "We now have binge-watching. The freedom not to wait a week for the next episode was huge. How much of TV’s formula is based on this?"

Despite this, Netflix still produces traditional episodes, although it, "Now skips the recap if you have watched a few in a row. It is possible to have a more human relationship when not driven by ad revenue," she says.

Neasmith said that not all producers want to move this way, and Matthew Weiner, producer of Mad Men wanted to maintain the weekly drip feed of content.

Editorial vs algorithmic

At Netflix, the recommendations are curated, but algorithmically driven. "When I got into this there was a war between Hollywood and Silicon Valley. Our CEO believed in an algorithmic approach, but we never got there. Content started to matter. So Netflix is curated, which shows balance, but it is highly algorithmically driven."

Algorithms can be a little too accurate when making recommendations. While an optimal appraisal of a viewer is more accurate, it can be a little disconcerting to the viewer. Sub-optimal choices seem to work better.

The future - automotive

"Automotive is the hottest topic in Silicon Valley - automotive data. Look at Tesla, Uber and self-driving cars." Neasmith said Tesla's approach, where the company learns about its drivers, is showing the car manufacturers the future of data.

"At the moment there is not much data in automotive, but this is starting to change. We need to get better mapping etc," she said.

Beyond automotive, we will have the Internet of Things and big data will just continue to grow. Now might be a good time to at least start thinking about how to process big data sets and at start learning more about your customers.

Getting data right balance between data and creativity makes for award-winning work. For tips on winning a DMA Award, come to one of our Awards Unplugged events, on 30 July and 13 August.