Skip to main content

Analytics inside: Why build, when you can buy?

A few years ago, you were ahead of the curve, seeing the power that data-driven decisions would bring, recruiting a [then] exotic team of super-smart and equally expensive data scientists, building some really cool stuff, and proving to everyone that it wasn’t just the titans of Silicon Valley who could transmute data into gold.

But now the rest of the world has caught up.

And these days, pretty much even the most mundane business software is claiming some analytics / machine learning / AI (choose your preferred jargon!) is already embedded. And in most cases – even allowing for the usual vendor hyperbole – they are right: off-the-shelf products are ‘smarter’ than they’ve ever been.

Are there any new worlds left to conquer?

Actually this is the wrong question – there is plenty of value still out there.

Be that in improving your existing algorithms, enhancing the data you use, enabling the decision-maker to more readily use these (e.g. never underestimate the impact of investing in better UI/UX/process-changes/workflow), or just addressing the business questions you’ve not yet got around to.

The real question is: where should one invest in being different, and where does one benefit from adopting something someone else has already created?

In short, now such sophisticated solutions are commonplace why not just buy everything as part of already proven, fully-formed, packages?

Standard can be great

Buying a pre-built tool should be cheaper, faster, lower risk, and will likely offer a complete, integrated, solution to a particular business problem.

For example, a commercially available recommendation engine might include various algorithms, but also connectors to the common marketing and e-commerce platforms to execute the recommendations easily, and a user-interface and workflow carefully designed for marketers or traders to be able to make adjustments at pace.

However, what you get is a highly ‘standardised’ solution which although you can probably configure to [some of] your needs it is unlikely to be tailored perfectly.

Of course, this isn’t going to be an issue if the process you are seeking to use it for is relatively standard.

The power of bespoke

Building your own solution in-house is pretty much the opposite. It gives you complete control: after all, who knows your business, and its data, better?

As with anything bespoke, this comes at a cost. You will likely never invest the same amount of resource into building such a rounded solution as a commercial provider who can, in effect, spread the development costs across a larger set of businesses (even before you start to add in the ongoing maintenance and support costs).

If you don’t need a fully formed package, with all the bells-and-whistles then building can be made to stay competitive with buying.

And in-house builds have a hidden super-power: for the right business problems, they can give you a chance to make your own algorithms a true differentiator from competitors. Otherwise, the only differentiation will be in your data and how you choose to use the solution (i.e. your business rules and your strategy).

Looking back to our earlier “recommendation engine” example, if you felt that the nature and quality of your recommendations could dramatically set you apart in the eyes of your customers, then you might be willing to invest to build something very different from what most competitors will achieve with their ‘standard’ off-the-shelf approach.

So, where does your business need something unique?

Unfortunately, the decision to build or to buy is never going to be quite that clear cut. Factors that will influence the decision include:

  1. Differentiation: as already discussed, if there is value in doing something genuinely different from your competitors AND that requires a unique algorithm (as opposed to data or strategy) to achieve this level of differentiation, then by definition you’re not going to get that in an off-the-shelf solution. The inverse is also true.
  2. Scale: crudely stated, when you consider the scale of return for a given use-case then only the biggest organisations may be able to afford complex builds. For smaller businesses, buying may become the only cost-effective option to get a robust, practical, solution that keeps up with the competition.
  3. Capacity: if at any point you happen to have ‘spare’ capacity in your data science and product engineering teams, then this can alter the cost equation in favour of building (remaining wary of ongoing maintenance implications, of course…).
  4. Capability: if the problem would benefit from very specialist knowledge (e.g. algorithms to model consumer pricing dynamics), then maintaining that rarely called-upon expertise in-house may be too expensive. In such cases buying a solution built by such experts can be more cost effective.
  5. Re-use: sometimes smaller components can be re-used as part of multiple bespoke solutions (e.g. a data matching algorithm). If you are likely to be able to re-use such components then this can tip the balance back towards “build”. This is more likely to have an impact if your overall strategy happens to be weighted to build vs buy.
  6. Roadmap: finally, it can be worth talking to the vendors of the major software packages that your business already uses. Are they planning on introducing some of the functionality you might be considering building? If so, might it fit your needs or can you help shape it?

Finding the right balance

A few years ago, you were ahead of the curve. Now you get to enjoy surfing the wave that has grown behind you. You are able to benefit from an increasing choice of cheaper, feature-rich, integrated solutions for the more common processes…leaving your team to focus more keenly on the areas where data and analytics can truly differentiate your business.

There’s plenty left to improve but isn’t it great that the rest of the world is catching up?

Meet the author

Peter Colthurst


Peter helps organisations to get the most from data including: building their in-house analytics capabilities; designing, developing, and purchasing, analytics solutions; and rolling these out to the decision makers who need to use them. His career in analytics began over 15 years ago running applied research programmes in the UK Public Sector to find early uses for then-emerging AI techniques such as Natural Language Processing and image analysis. These days he works primarily in the Private Sector, aiding consumer-facing businesses in the UK and elsewhere around the world.

Sign up for the latest updates

Catch up on Experience Analytics:

The Virtual Series - watch the episodes you missed or re-watch your favourites!