Reimagining the Future with a 360 perspective

The text below is an extract from Jules Goddard and Rob James’s recent book on business experimentation

Developing the hypothesis

What is a hypothesis?

In many ways, a hypothesis might be seen as a prediction, but it is much more than a guess; it is informed by a deep understanding of the problem or opportunity that is followed by an exploration of potential solutions.

A hypothesis then is ‘a precise, testable statement that proposes a solution to a problem where the specific outcome is anticipated but uncertain’.

A testable hypothesis is a conjecture that can be proved or disproved as a result of experimentation, data collection or experience. It may be considered a reflection of a team’s expectations, but the point of the experiment is to determine whether the statement is true or not. A disproven hypothesis can be just as valuable if it provides robust data on which future decisions can be based.

In a business experiment, teams often fall into the trap of creating a hypothesis early in the process and rarely referring to it again. The hypothesis is one of the most important tools of an experiment. It should be a constant reference point. It is fundamental to how the experiment will be set up, how its progress is monitored and how the results are evaluated. In a car it is the equivalent of the satellite navigation providing direction, the speedometer, odometer and fuel gauge monitoring the journey, and the warning systems alerting the driver of potential hazards.

Writing a hypothesis

We ask clients to write their hypotheses in a relatively simple way. By this stage they are already very familiar with the problem, so we concentrate on the possible solution and expected outcomes in the following format:

If we do X, then we will achieve Y.

X is the independent variable. This is the selected single intervention or the element that will be changed for the experiment. It is assumed to have a direct effect on the outcome – the dependent variable.

Y is the dependent variable and is the expected outcome that will be measured.

We describe two examples of hypotheses in the examples below and, while they may appear to be uncomplicated statements, they are more intricate in their construction. We expect the hypothesis statement to be a portal that uncovers exactly what is happening in an experiment.

Example 1

Problem. An international recruitment firm based in Australia recognized a lack of ‘strategic’ oversight with regard to their largest client, a global online retailer. they were operating out of four offices, separately located, each one serving a separate part of the client’s business. they were missing synergies and this was resulting in poor coordination and reduced productivity. with each location managing its own client relationships, it also restricted opportunities for cross-selling. the recruitment firm preferred a ‘partnership’ arrangement with key clients as they considered it better value for the client, but relationships were not at a senior level and most interactions were transactional. the team researched the problem both internally and with the client and developed this initial hypothesis:

If we create a centralized account management system for client A, then we will improve strategic oversight.

This works fine as a headline hypothesis but prompts many questions. It demands a deeper level of interrogation. when the team were asked to be more specific about the outcomes, their hypothesis was expanded. As an example, and taking the dependent variable of ‘strategic oversight’, the Y part of the hypothesis became:

If we create a centralized account management system for client A, then we will enhance client satisfaction, improve productivity, and increase EBITA.

This only a provides a part of the interrogation as each development of the hypothesis should raise another level of questions: what do we mean by ‘productivity’? do we have base measures in place to evaluate the changes in client satisfaction? By how much do we expect to increase EBITA?

The same type of interrogation is appropriate for the X element (the intervention) of the hypothesis: what do we really mean by a centralized account management system? How do we fit this with existing systems? As the interrogation of the hypothesis continues, it produces the basis of the experiment plan.

Example 2

Problem. this chemical firm based in Germany was finding it increasingly difficult to attract graduates in a diminishing talent pool. there were particular shortages of engineering candidates and a study indicated that over 90 per cent of those interviewed did not see the company as a ‘preferred employer’. the company was using a standard job template that was not tailored to specific professions or target groups. the experiment team decided to try a relatively simple experiment. Based on the feedback from their research of the problem, they produced a small number of tailored advertisements to compare with the standard advertisement (as the control) using the same job portal. their hypothesis was:

By tailoring job adverts to the needs of specific target groups we will increase the attraction rate and the number and quality of applications.

In each hypothesis statement the underlined text is X and the text in bold is Y.

Extract from “Business Experimentation” A practical guide to driving innovation and performance

Authors – Rob James and Jules Goddard

 © Rob James and Jules Goddard 2021