Analytics is not visualisation. Most of our customers start off wanting better analytics but they really mean visualisations.
For example, we often engage a new customer by building them some demonstration dashboards and reports. It’s about 10% of the platform capability, but it’s an important 10% that they can see and relate to, particularly as we do it using their data.
To make things easy we have Sabisu Publisher. We get the customer to send us some representative data in MS Excel and we upload it, checking the ‘Publish as data’ checkbox, which decomposes it to raw data and puts it in a database.
Then we build some simple views to get some feedback so we can better understand the customer’s requirements:
This is not analytics. This is simple visualisations of data. Often you could do the same in MS Excel but you’d be missing:
- Integrated click throughs & drill-downs
- Easy collaboration (chat, conversations, posts)
- A single, shared, trusted report for all
- Automatic historisation for building lots of different reports quickly
At this stage, most customers are delighted. They have a smart dashboard which they can evaluate, present to their leadership team and use as a vision of the future.
Analytics it isn’t.
Analytics is the processing of the data to generate new insight or elucidate that which was hidden. All the smart visualisations in the world aren’t going to help you if the correlations or relationships remain hidden.
To properly implement analytics takes time and expertise. We’re fortunate in the oil & gas/petrochems sector that the challenges each customer faces are common to the sector, e.g., reporting OEE, alarm analysis, asset health monitoring and so on. This means we can get a head start in using complex analytics by preparing for common requirements.
Here’s an example of using statistical cluster analysis to identify alarm system redundancy. Having lots of alarms occurring is a bad thing and is often identified as a contributory factor to major oil & gas and petrochemical incidents. There are guidelines (EEMUA191) to adhere to, so if you have a noisy alarm system, you want to identify those that are redundant and eliminate them so your operators see only alarms which are important.
Using some smart mathematics to do the analysis, we can easily call out clusters of alarms which act together. Redundancy is obvious:
Those visualisations are smart; force directed graphs and matrices. Nice. But it’s the analytics behind it that is doing the heavy lifting.
For example, these two tables show all alarms ranked using NodeRank and PageRank. The clever bit is the ranking; the tables are just…well tables:
Not sexy. But very valuable, because I now can which alarms ‘triggering’ the biggest cascades and being statistically ‘fired’ by other alarms the most. Now an alarm management change control process can kick in, taking redundancy out of the system to make the plant safer.
So, a simple table might be the most effective analytics visualisation of all.