Every customer engagement is different. It’s what makes life so interesting – and, sometimes, challenging.
One of those challenges is that while the customer wants to move towards intelligent and predictive analytics, they’re just not ready for a full implementation. This can show in many ways, including:
- Insufficient dataset size or quality for deep analytics or algorithms to reliably work.
- Not having a clear idea as to what questions need answering – in other words, not knowing what you want to know.
- Not having the appropriate context around the data, so it’s not clear when the data indicates ‘good’ or ‘bad’ performance.
The starting point for us is that the customer just needs help understanding where they are and how they might derive more value from the data they already have. Practically this means engaging with the end-user to get something useful now that will build over time to be a valuable resource that can be interrogated by algorithms or analytics to greater effect.
Build algorithms and calculations using existing systems and available data.
Whilst not deep analytics, we can start to build a large dataset by interrogating existing systems in real-time and carrying out linear or non-linear calculations. For example, a compressor may have key data reported every couple of minutes; we can take that data, apply some non-linear physical properties modelling for steam and compressor performance and derive a much richer dataset.
Build useful visualisations.
Algorithms are great with vast quantities of data; humans are great at pattern matching and recognising relationships. If we have limited data then we try to use interesting visualisations which will allow users to spot these relationships. If the end-user can then record these relationships, adding metadata, then so much the better. That’ll help us later.
Educate users on collaborative capabilities.
Which brings us to metadata capture. End-users have a wealth of knowledge, untapped and untappable by the algorithms we will eventually want to use. We drive users towards the collaborative capabilities of Sabisu, encouraging the recording of tacit knowledge. Short term, this promotes sharing of best practice. Long term, this gives context to the data which makes it more useful to the algorithms when they’re eventually deployed.