Andreas Soller

Double Diamond Model


Reading time of this article:

3 min read (731 words)

Publishing date of this article:

Apr 28, 2024 – Updated May 1, 2024, 08:09

This article is tagged as:



In 1996 Béla H. Bánáthy proposed the divergence-convergence model in his book “Designing Social Systems in a Changing World”. This design process model became popular when it was enhanced and published by the British Design Council as Double Diamond model in 2005.

  • Divergent thinking refers to exploring and building up a deeper understanding: create options.
  • Convergent thinking refers to having fully understood the options and hence, to be able to make a choice.


Discover refers to all activities to get a deep understanding of the problem.

In this first phase you might together

  • observe and
  • talk to users,
  • execute desk research to understand how this problems are approached in other contexts,
  • you will run surveys and
  • check quantitative data,
  • map different context situations over time,
  • etc.

As of research triangulation is key: that means you will combine different research methods such as interviews, observations and quantitative data to get a fuller picture.


In this phase, the team will get a deep and shared understanding of the problem space.


Define refers to all activities to come up with a first problem statement.

You will analyse your findings together

  • and create personas to get a better understanding of user segmentations,
  • use frameworks such as jobs-to-be-done to understand the tasks
  • use mapping techniques such as User / Customer Journey Mapping for the different user groups (personas),
  • create a service blueprint in case you want to analyse the User Journey together with internal processes,
  • etc.

The goal is that you come up with a single problem statement:

GIVEN THAT (context, situation)
HOW MIGHT WE (question)
SO THAT (goal, objective)
BECAUSE (need)

A good problem statement is

  • specific enough to address the problem it wants to tackle,
  • broad enough that it doesn't specify the solution upfront,
  • has a multifunctional perspective and
  • inspiring to the participants that they want to solve it.


In this phase, the team will come up with a single problem statement as starting point to explore possible solutions.


Develop (design) refers to exploring potential solutions.

In this phase you will ideate together and

  • phrase refined hypotheses,
  • create storyboards and
  • rapid prototypes
  • to validate your ideas.

Rapid prototyping refers to shitty first drafts that you can you to extend your learning about possible solutions. You don't create high fidelity prototypes but rather quick prototypes to get fast feedback. A rapid paper prototype will also give you more insights as long as you are exploring, as first test users will more easily tell you what is still missing or what can be changed.

Don't focus on specific ideas but rather go for different ideas. Think holistic and run rather full service concept pilots instead of very specific solution assumptions.


In this phase the team will have further enlarged their problem understanding by exploring various ideation techniques. Additionally, the team has iterated and validated various ideas and build up knowledge about the impact of various solution assumptions.


Deliver refers to all activities where you evaluate different solution assumptions.

You will together

  • define criteria
  • to select options,
  • check impact objectives
  • do final validations,
  • come up with action plans,
  • and a concrete solution assumption
  • etc.


In the last phase the team will have come up with a concrete solution assumption – the next best test (MVP).


Most important, Design Thinking is not a one-way-street but rather an iterative process where continuous touch points with the users are key. It is not enough just to create some output. Instead you focus on outcomes for the users that will use your product or service. The process requires a constant reality check and continuous learning.

You just deliver the smallest output that is needed to test your assumption with a large enough sample to gather enough confidence for next steps. With new learnings you will either move on or iterate each of the four phases.

It can also happen that the outcome of one loop is a readjustment of the initial problem assumption. In this situation you iterate the full loop once more.

References and further reading material:

How did you like this article?

Thank you for your feedback.