Buying Analytics – a rough guide

This is not my Jerry McGuire moment. I’m not trying to come clean and reveal the dark arts of the IT salesperson. In my experience the more open and even handed you are, the more trusted and the more successful you will be. Instead, I thought it may prove useful to shatter some myths and shed some light on a market that is often over complicated – Advanced Analytics. This is aimed at the buyer and user of analytics solutions. I hope it helps.

To be fair, it is easy to see why it has become so opaque. Businesses, organisations and even small departments recognise the value of getting answers and insights from their data. The key challenge is that the volume of this data is growing hugely, customer, sales, competitor data, logistics, personal data, IoT and device data, cyber security logs, the list is endless. Then add to this sector and domain specific data. In my world – pharmaceuticals and life sciences – you can include longitudinal patient data, clinical trials, drug safety, regulatory, channel sales and much, much more.

So, we have a lot of data to handle. Well that’s okay because we now have data scientists, Spark, data architects, Kylo, analytics consultants, analytics in the Cloud, Hadoop…I’ll stop before you get a headache. It is a personal frustration that many vendors seem to delight in this tech talk and the challenges end users face.

My first warning is to beware of the IT vendor who only talks technology. It should be about the business and the people that make up your business. The first question out of you vendor’s mouth should be something like; “What is your business trying to achieve?” Or a more nuanced pharma challenge might be; “I’d like to run even more real world data queries to help design my clinical trials but each one takes so long that the compound effect is untenable.”

The vendor should be talking about helping you discover the root cause of the business problem, simplifying and speeding up your analytics or simply getting your data out of silos and making it available to the whole organisation.

Beware the vendor who tells you to get rid of everything you have and then wants to sell you a whole stack of applications and tools. For a start any decent technology should be relatively interoperable with your existing set-up. Obviously, some technology is so ‘legacy’ that you really should consider the museum or landfill.

In my experience customers can also believe the hype.  I have met the ‘I’ll consider anything as long as it’s Oracle’ guy, or the ‘We’re replacing Netezza and looking at Snowflake and Redshift’ team. Never mind that by many measures they are not even in the Top 5 Cloud Analytics solutions. See 2018 Gartner Critical Capabilities for Data Management Solutions for Analytics

On the subject of the Cloud, it’s worth mentioning that the Cloud is excellent, but not all Cloud offerings are equal. Once you have established how easy it is to get your data into the Cloud you might want to ask some more detailed questions around performance in the Cloud and indeed how easy and costly it is to get at your data.

In this vein we occasionally get the challenge that the customer is thinking about moving everything to Hadoop and replacing their existing data warehouses. Our response is invariably words to effect of; “Sure, just make sure you understand the ease of getting data back out and the performance levels of analytics on Hadoop.” Hadoop can be great for ‘cold’ data, archiving and generally storing data that you don’t to work with frequently and quickly.

At Teradata we talk in terms of architecture, our recommended one – the Teradata Unified Data Architecture – as well as the customer’s existing one. A customer of mine referred to a Standard Data Pattern. Here the best solution is invariably one that builds upon and enhances what the customer already has. In this scenario Teradata becomes the analytics platform and we sit very comfortably alongside SAS, Jupyter, S3, Apache, R3, Python and more.

Beware of the salesperson who presents an us versus them scenario, or rather us versus all the rest. Just as there are lies, damned lies and statistics, there are also some equally wild claims.

We recently had a customer come to us and say that an up and coming Cloud Analytics company was aggressively coming after new business and was offering a challenge in the shape of a Proof of Concept (POC) test. Specifically, they were saying something like; “Give us your four most complex queries and we’ll show you how good/fast we are.”

We were asked if we would like to take part and we politely declined. They tried the POC and signed a deal assigning some of their data to the new vendor. They then returned to Teradata after six months saying it hadn’t worked out and asking us, out of interest, why we had declined the test in the first place?

Here’s why. Teradata’s data scientists and consultants thought the POC was pointless from the outset. Any analytics software should be able to handle a complex query. What separate the grown-ups from the rest is whether or not the engine can handle high numbers of concurrent queries, spikes and high volumes of users and large and complex data sets. It’s not the one, it’s the one hundred or thousand that counts.

So in summary, focus on your business and your users. Appreciate what your data scientists like using and ask yourself how you can make them more productive. Clinical data scientists in pharma R&D tend to use SAS and invariably will have been doing so for years and years.  So rather than taking away their toys, you might want to ask what the benefits of running their SAS queries on another platform as opposed to SAS on SAS.  We’ve partnered with SAS to create a high-performance offering, the outcome of which and evidenced by numerous benchmarks, show that SAS running on Teradata expedited the data manipulation 500 to 1000 times faster than SAS only.

So, don’t get side-tracked and blinded by the latest tech hype and make sure your IT vendor salesperson is talking about your business and your people first. Better yet if they are big enough to respect your current set-up and even talk about working alongside other solutions.