Why does data-driven decision making fail?

As increasing amounts of data become available, businesses that are effectively harnessing it are pulling away from competitors. The digital universe is exploding, indeed by 2025, it is estimated that there will have been a 50 fold increase in online data available within just ten years. It is well known that data encourage well-informed decision making. Especially in a situation where there are various stakeholders with conflicting perspectives, hard data can act as the mediator, ensuring that business resources are allocated optimally.

However no strategy is infallible, and making decisions based on data does not always lead to optimal outcomes. Reasons include:

Mistaking correlation with causation and conclusion bias

Two events happening at the same time may or may not be linked where one may or may not be causing the other. There is a risk that superficial analysis can lead to conclusion bias. That is, the person doing the analysis has a personal opinion which they look to validate through the data. One of the most famous examples of this is the correlation between psychiatric disorders and recreational drug use. Is it the case that drugs cause disorders, or that people predisposed to disorders are more likely to take drugs?

Whilst there is a correlation, causation either way is not proven. Within the job market, an example of correlation over causation would be, the economy is growing and more people are employed. This is a great example of why diversity in the workforce is so important, as conclusion bias can be directly attributed to upbringing, socioeconomic status and peer groups. Having a diverse workforce neutralises this. With all this in mind, if your business has a culture of data-led decision making, having multiple people with different perspectives analyse the same data sets will give optimal results. An excellent book to read on this topic is the Black Swan.

Failure to implement insight-led outcomes

As data is integrated from various sources, the complexity can become overwhelming. At that point, there is a danger of paralysis by analysis. With so many variables, it becomes unclear what the correct course of action is, where instead of data being used to drive change nothing happens. Therefore at every step, there needs to be clarity on what the objectives are and how the data on hand can help answer that question. For example, now companies house allows for API integration, what this means is that recruiters could have the financial information about all their medium / large clients contained within their CRM.

That sounds impressive, but unless there is a clear outcome in mind, which will generate ROI, so what? So why not then create reporting that shows for all the key clients, what their year on year turnover and revenue growth is, so to help with resource allocation for key account management? i.e. with faster-growing clients allocate more resources, with loss-making clients, assign less.

Relying on low quality or inconsistent data

Data led decision making is dependent on the quality and consistent data. As an example, imagine you are analysing a client’s hiring activity over the past year as part of an annual review, but six months data is missing. Or, when you analyse hiring by role, jobs are not categorised consistently. Doing any kind of meaningful analysis becomes impossible. You are better off not trying, as the alternative is to make assumptions that could end up being faulty.

At Vacancysoft this is why we have strived to create a data set that is both high quality and consistent. This starts with our company coverage. In the UK, medium to large companies account for over two-thirds of all private sector employment, so our starting point is to monitor them comprehensively. Whenever they republish a job, we know. If they change their career centre, we track the new one. This means that with your key accounts, we are able to provide consistent coverage.

In the same way, our specialist data quality control process means that every posting is categorised using our proprietary taxonomy, so results in standardisation of job processing, meaning that you can analyse activity by client or market segment, without having to worry about false readings due to low-quality data.

If you would like to see how your decision making could be enhanced through vacancy data, contact us for a free consultation.

About the author: James Chaplin
Tell us something about yourself.

Get involved!


No comments yet