Arrow DownArrow ForwardChevron DownDownload facebookGroup 2 Copy 4Created with Sketch. linkedinCombined-Shape mailGroup 4Created with Sketch. ShapeCreated with Sketch. twitteryoutube

A new series of Grand Designs started on TV recently. The show takes you on the journey a couple go through when designing and building their own home. You see some incredible architectural achievements and beautiful buildings. What the show reveals is that, despite the enduring romance of self-build, building your own home typically takes longer, costs more and ages you.


Mojo Mortgages analysed all 124 episodes to reveal that 80% of the self builds went over budget by an average of £124,000 (34%). Little wonder then that building your own home is not as popular as buying one.

When looking at the world of software I see a similar story. It’s still possible to develop a word processor from the ground-up, but few would choose to. In 1999, Goldman Sachs stopped using the word programme it built for itself and integrated Microsoft Word into the workflow for example. It’s a similar story for firms that took a self-build route with their CRM systems, or ecommerce platforms in the early 2000s.

I believe that data science will be no different. Most data science teams I speak to today have cobbled together their own set of tools for doing their job. But the value that a data scientist brings to a company by actually doing data science will always be greater than the value of the time they have to spend building and maintaining their own set of tools for doing data science. Tasks such as spinning up development environments, sharing data and code, scheduling and running jobs, tracking experiments and productionising models.

So, is it a good idea to meet data scientist needs via grand design or buy something instead? Here are two occasions where we think it’s the right decision to invest in your own bespoke Grand Design:

  • When there is certainty that it will be fast and easy to build using existing resources.
  • When the system is for the core processes that differentiate your company.

In all other circumstances I would take the best offering from someone else, and here is why:

  • Speed: Buying a data science workbench as something that you can install today is a much faster path to business impact and is infinitely quicker than building one yourself. 
  • Simplicity: If it’s complex to build then expertise, maturity and economies of scale come from acquired packages. Most software projects fail – don’t take the risk.
  • Cost: The manpower it takes to build a data science platform is an order of magnitude more than buying something off-the-shelf. And, it’s easy to forget that 70% of software costs occur after implementation.
  • Quality: The features, support, and future improvements  will never be as good as the best off-the-shelf solution. 

So, if you’re reading this and you know data science could be done better in your company, what could you do? Here are three suggested steps. First – talk to the users! Data scientists are usually best placed to explain the importance of a high quality data science workbench they can use. Second, talk to other companies about their own experiences – for example our website lists some of the companies currently enjoying Faculty Platform. Finally, create a scored and weighted matrix with requirements or challenges you feel you experience on one axis and products that might solve them on the other. Identify the best solution for you, and get in touch with us!

To find out more about our data science platform or request a demo, get in touch: workwithus@faculty.ai

 

To find out more about what Faculty can do for you and your organisation, get in touch.