Omni Core · Personalization · Module

Whether it works

Evidence that the personalisation improves something for the customer as well as for the organisation, tested rather than assumed.

The idea

How it works

Personalisation is assumed to work because the logic is obvious. Tested against a non-personalised control, a substantial proportion of it produces no measurable difference.

The test also has to include a customer-side measure. Personalisation that raises conversion and lowers satisfaction is a trade rather than an improvement, and it should be made knowingly.

Working with it

In practice

  1. 01

    Hold back a non-personalised control

    Permanently, not just at launch. It is the only way to know what the personalisation is contributing.

  2. 02

    Measure the customer side too

    Satisfaction, effort or repeat behaviour. Conversion alone cannot distinguish helpful from manipulative.

  3. 03

    Retest as the model ages

    Personalisation degrades as behaviour changes, and degradation is silent.

  4. 04

    Remove what does not earn its maintenance

    Untested and unmeasured personalisation is a permanent cost with no established benefit.

One level in

The components of whether it works

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

  1. The control

    The non-personalised group held back for comparison.

    Learn
  2. The customer measure

    Whether it improved anything for the customer, not only for conversion.

    Learn
  3. The decision

    What was kept, changed or removed as a result.

    Learn

Personalisation that raises conversion and lowers satisfaction is a trade. Make it knowingly or not at all.