PAW: The unrealized power of data

Live from Predictive Analytics World

Andreas Weigend, former amazon.com Chief Scientist, gave a keynote on the unrealized power of data. He started with a historical perspective. In the 70s perhaps 10M used computers, mostly in the back office. By the 80s this had reached 100M and the front office. By the 90s the internet and search brought 1Bn poking around and some customer-company interaction. Now there are perhaps 100M producing content on the web in peer-production and collaboration – customers are interacting with customers. Underlying all this is a drop in communication costs essentially to zero. Now people can contribute and fix data rather than simply consume it and the time to respond – the natural timescale – has disappeared.

Some trends:

• There is now about 100Gb stored per person on the planet and it is doubling every year.
• Market research can now combine explicit survey data with implicit behavior data
• There is a move from models being assumption heavy to being data rich thanks to the number of visitors and the amount of information.
• From knowing about transactions (enough for recommendation) to knowing interactions (enough for targeting) and ultimately relationships (can…

Live from Predictive Analytics World

Andreas Weigend, former amazon.com Chief Scientist, gave a keynote on the unrealized power of data. He started with a historical perspective. In the 70s perhaps 10M used computers, mostly in the back office. By the 80s this had reached 100M and the front office. By the 90s the internet and search brought 1Bn poking around and some customer-company interaction. Now there are perhaps 100M producing content on the web in peer-production and collaboration – customers are interacting with customers. Underlying all this is a drop in communication costs essentially to zero. Now people can contribute and fix data rather than simply consume it and the time to respond – the natural timescale – has disappeared.

Some trends:

• There is now about 100Gb stored per person on the planet and it is doubling every year.
• Market research can now combine explicit survey data with implicit behavior data
• There is a move from models being assumption heavy to being data rich thanks to the number of visitors and the amount of information.
• From knowing about transactions (enough for recommendation) to knowing interactions (enough for targeting) and ultimately relationships (can move to a long term relationship basis).

The customer data revolution has led companies to “sniff the digital exhaust” and there is far more implicit data like location. In addition, individuals like to talk about themselves creating more data and they reveal their relationships with others in all sorts of way. But to get this information, and thus be able to use it, companies have to have a consumer-centric point of view. They have to offer consumers something in return for their information.

Andreas talked about moving from Customer Relationship Management to Customer Managed Relationships. True customer-centricity empowers customers to make the best decisions they can. Customer value is one thing – what is this customer worth to a company – and companies have a value to a customer. Needs to become a bi-directional relationship.

Companies no longer “own” the customer – customers are more likely to evaluate multiple companies online, for instance. Companies don’t know more about their products any more – the web does – and even cannot control their message or branding.

Marketing 2.0 is different:

• Communication is not just about companies targeting customers 1:1 but recognizing that customers communicate with each other 1:1.
• Customers like to review products before they buy them and prefer peer reviews. 
• Relationships also trump many other things. For instance marketing a phone product to those who were called by people who already owned it (using the relationships therefore of existing customers) outperformed a sophisticated marketing model by nearly 5:1. Network-based marketing or leveraging the social graph.
• Have added all sorts of information about friends, peers, expert bloggers, annotations and more. Using this requires new approaches.

He outlined a five step approach to applying this thinking – PHAME – Problem, Hypothesis, Action, Metrics, Experiment:

• Problem – defining the problem is key as many businesses have a problem different from what they think they have.
• Hypothesis – come up with a hypothesis for a solution. This, to some extent, relies on a culture of experimentation.
• Action – define the actions you are going to try in support of this hypothesis.
• Metrics – spend some real time defining metrics and measures that will both show that something works and that will encourage movement in the direction you want.
• Experiments – see what works, doing experiments is both expensive and yet it is cheaper than ignorance.

In conclusion he emphasized that communication costs falling to zero brings customers into the network but only if they get something back and only if the company respects the cost of their attention. Using relationships can result in dramatic results if a experimental and metric-driven culture can be created.

More posts and a white paper on predictive analytics and decision management at decisionmanagementsolutions.com/paw