Omni Core · Personalization · Module

The signals

What the personalisation is based on — stated preferences, observed behaviour, inferred attributes — and how reliable each is.

The idea

How it works

Stated preference is accurate and rare; observed behaviour is abundant and ambiguous; inferred attributes are abundant and frequently wrong. Most personalisation runs on the third.

A wrong inference is experienced as being misunderstood, which is worse than not being personalised at all. Error rates matter more here than in most data applications.

Working with it

In practice

  1. 01

    Prefer stated preference where it exists

    Asking is more accurate than inferring and is frequently possible.

  2. 02

    Know the error rate of each inference

    An attribute inferred with seventy per cent accuracy is wrong for three people in ten, permanently.

  3. 03

    Let people correct it

    A visible and correctable inference is tolerated; an invisible wrong one is not.

  4. 04

    Check the signal is still valid

    Behaviour from two years ago describes a former customer, and stale signals produce confidently wrong personalisation.

One level in

The components of the signals

A component is something that exists afterwards which did not exist before — a deliverable or a mechanism, not an intention.

  1. The signal set

    What each personalisation is based on.

    Learn
  2. The accuracy

    How often each inference is right.

    Learn
  3. The correction

    How a customer can fix a wrong inference.

    Learn

An attribute inferred at seventy per cent accuracy is wrong for three people in ten, permanently.