Small and agile or large and powerful: size matters

Businesses must ask themselves, where does data science fit?

There has been much hype around data science, and in particular recently about artificial intelligence (AI), with Google’s DeepMind announcing that its programme AlphaGo is able to “create knowledge itself”, and Accenture Research and Frontier Economics suggesting that AI can enable 38 per cent profit gains by 2035. All businesses would like to make profit gains, and many are beginning to derive value from data science, but they are facing a new challenge: where does this new capability fit in their company?

One option is a central data science team tasked with supporting the whole company. Another is several smaller teams that work with specific business units. Another is to outsource data science. And of course, it is possible to implement some combination of these approaches.

A centralised data science team can ensure the benefits of new tools are widely shared, and that work is not being duplicated in different parts of the business. As a focal point for data science, it can help to make the potential applications and benefits visible to all. And a central team may be large enough to justify investment in training and infrastructure that will make it more productive than smaller teams.

However, a centralised data science team can run into problems. It may not be able to respond quickly to the needs of different business units. It may standardise on technologies which do not fit with the needs of some parts of the business, slowing down solution development and frustrating staff. Centralisation and standardisation have many benefits, but a delay in getting that first application into production is particularly an issue for data science, which for many companies is a new and unproven capability which needs to demonstrate its value. Smaller teams that work within business units can be more flexible and will have a better understanding of their unit’s requirements, which can help deliver better results.

There is, therefore, a trade-off between creating a centralised data science team which will be large enough to develop significant technical expertise, or smaller more flexible teams within different business units which will better understand the domains in which they work.
Outsourcing data science can give companies access to a pool of staff larger than even the biggest internal data science teams, who will therefore have the right specialist skills and abilities for a wide range of tasks. An organisation with such a large pool of data scientists is also well placed to provide the mentoring, training and career development that leads to significant technical expertise.

The wrong outsourcing model can exacerbate the problems of the centralised team, with the outsourced data scientists invisible and inaccessible to staff in the business, but the right outsourcing model can help overcome this and other problems. Embedding at least some of the outsourced staff within the company helps visibility and access, and the scale of a dedicated data science organisation means they are likely to have the varied skills necessary to meet the needs of many different business units. The ideal situation is to have the data scientists easily accessible to all parts of the business while keeping a firm control on costs. There are many ways to do this, but the best ways require a level of trust, which means that this is a relationship that both sides should be viewing as a long-term investment.

There are therefore several options if a business wants the profit gains Accenture suggests, but the right option must be chosen quickly if the gains are to be realised. Companies starting with data science should begin to see useful data come through within weeks and actionable insights should be available within one year, maximum. If they continue to persist with a platform or approach, that is not delivering value the company will not only lose money but could also see relationships sour and staff members lose faith in data science. This could mean that businesses could be missing out on transformational capabilities because they hadn’t planned their previous attempts well enough.

Data science began its business life much like a cottage industry, with staff members interested in the field exploring it on a small-scale. If organisations are ever going to get to a point where they can deploy and institutionalise data science to deliver sustainable value, decision makers must think carefully about where data science fits in their company.

 

Categorised:

Tim Pattenden

Tim Pattenden

After university, a short stay in a City accountancy firm convinced me that what I really should be ...

© Copyright 2018 Tessella
All rights reserved