Friday, April 10, 2009

Capturing Intention in User Behaviour - Do it Directly

Taken from Princes St.Image via Wikipedia

I had an interesting conversation at OpenCoffee London yesterday around personalisation and recommendations. At they table were several people involved in companies that are in the area of marketing analytics and personalisation and recommendations.

One interesting point raised was that as targeting becomes more and more focused, there is a loss of information that can harm personalisation/recommendation algorithms. There is a point where targeting stops being effective in delivering relevant personalisation and recommendations. I'll return to this later.

Of more note was how capturing attention is difficult, indeed impossible with behavioural tracking. In "Wither Social Networks, Arise Communities" and "Resources vs Answers - Asking a Question of Search" posts I've look at the importance of intent to properly target a user. What came in the conversation yesterday was be direct. Essentially, ask the user what their intent is and stop trying to guess. Allow the user to tell you (the business/software etc.) exactly what their intent is.

The example used was booking flights to Edinburgh. Currently travel sites get you to input the departure, destination and dates. Then give you lots of results which you have to wade through and links to ads for various travel related items in Edinburgh. However, there is no way for the site to tell what links it should show or how to order the results.

If however, the user starts by telling the service "I want flights to Edinburgh for a weekend of fun" suddenly the service has a slew of information that it can use to target the right ads and services and also organise the results. From that simple statement, the system "knows" the following information:
  • The flights need to be Friday afternoon/evening and Sunday afternoon/evening
  • The purpose is fun so flights need to be cheap/reasonable
  • The person will be looking for tourist, entertainment stuff to do
From this new knowledge the system can prioritise flight results and target service suggestions and advertising better to the person's intent. This wouldn't be hard to do. It is a step away from normal form-based UI into NLP but it opens a whole new vista of opportunities.
Reblog this post [with Zemanta]