Moving from this world to a data management system can do several things quickly. A company can get access to this information in real time, and because scientists are not spending as much time sorting through data, its research capability will increase. There can be other benefits, too. Research efficiency could improve because researchers will not have to redo work done previously by other colleagues because they cannot find the data. They can more easily analyse data generated by their colleagues, draw conclusions, and develop solutions.
A good data system also captures expertise so that the knowledge of very experienced people is not lost when they retire. This information will also be in context, so new starters have fewer questions, and should be able to get up to speed more quickly. Such systems can also make it possible to capture more data, to fill the information holes, and give better modelling and information tools.
The polyurethane industry produces materials as diverse as engineered foams, adhesives, coatings, rubber and rigid foams; while the data which R&D, technical service, and production generate are structurally very different, the companies innovating with polyurethane have a similar problem. At its core, the question is ‘what do the data really look like?’
Polyurethane formulations are relatively complex. Each raw material has its own physical and chemical properties, and these are not completely consistent batch to batch. As well as physical and chemical properties, raw materials have other attributes that need to be considered, including their price, and whether they are affected by regulatory controls.
Once the reactive chemicals are formulated, new layers of complexity emerge during processing, around mixing, curing, heating and cooling, for example. Once a sample is made and tests are carried out, more types of data emerge. These could be physical measurements or images or curves, or numbers from field testing. All of this information needs to be pulled together for each formulation, and for each experiment.
Many people pull the results of their experiments together in spreadsheets. These are unstructured and flexible, and they can perform calculations and can be modified from user to user. This is a strength and a weakness. Some teams use freeform electronic lab notebooks with purposes and hypotheses as well as data, tables and, possibly, images to be captured.
Both spreadsheets and electronic notebooks are frequently used to track product requests and trials, and collect the results from tests. Some technical departments use also use predictive tools.
Away from the lab, companies also have inventory tracking, ERP and regulatory compliance systems. There is often a wealth of structured information, which may not be immediately accessible to the technical teams or tied to data generated by the R&D or technical departments.
But there is a problem. Spreadsheets, documents, pictures and pdf and lab journals are all unstructured. They don’t force users into a syntax or way of operating. For large operations running many different experiments in different labs, unstructured information and systems very quickly become highly complex and difficult to manage.
Give it structure
By keeping information in an unstructured form without its own syntax and rules of operation, companies lose the ability to mine the information they generate in a scalable way. Their knowledge can be limited to a simple keyword search in each of the silos where the information is stored. Each search will return its result in isolation, without related data. This level of opacity and complexity can make it very hard for companies to answer a simple question: how long would it take you to find out if you had data that showed a product met a customer’s criteria if that information in in data generated by a colleague?
But, as we have seen, not all the information within companies is unstructured. For example, ERP and CRM systems are structured data environments. Uncountable system is in this camp, and can provide instant transfer between these systems. With the correct permissions, scientists are able collaborate more effectively with their colleagues. This can help them to find the best starting point for a customer specification within a couple of minutes, rather than a couple of hours.
We recognise the benefits of structured data and will work with companies to ensure that they find using the system very easy. The bigger challenge can be to get technologists and scientists to stop using simple spreadsheets.
One of the reasons that people in the polyurethane industry like to use non-structured information is because the results that tests produce are noisy. Polyurethane systems have to be tested in real life, and this means results vary and are not completely reproducible. For example, tests on a polyurethane sample may give different levels of variability between different properties; density results may be closely grouped, but compression set results could be more widely spread.
It is a multi-layer problem. How do the properties of raw materials and their functionality affect the properties of the finished foam or the processing properties of the formulation? The information generated in an experiment could be of several different forms, they could be a time series, temperature or rate dependent for example. Data points may not be dense enough to use big data processes to analyse data. A data capture and recording system has to be able to accommodate these variations.
Another problem the polyurethane industry faces is that many senior scientists have been working within a company for decades. They have a great deal of knowledge, and it can be challenge capture their knowledge before they retire.
Before implementing a data capture system, companies also need to think about the business tools that they want to join up. Do they include ERP? Or regulatory? Finally, who should have access to the data? Where will the data be audited and stored?
These are usually the main areas of friction that companies and their R&D teams need to overcome when they adopt a more centralised system. We find that the best way to get people on board with the change is start small and look for quick wins. If we tell scientists, ‘do all this work here and then you’ll see benefit in 12 months or 18 months!’, very few are going to be motivated.
If they can see the benefit within an hour or less, that helps. It might be comparing, say, DSC or analytical data very easily. Enabling scientists to find a couple of ways to save a ton of time very early on will incentivise them to go back into the system and input more data.
Keep it new
Old data are, of course, important, but the projects that are being worked on are immediately relevant. When making the change, it is important to think about future information needs. We advise that a company should focus on those data that are important right now, and do not make the addition of old data the guiding factor, especially if it takes a long time to implement. Put all the current and future work into the system and, when time allows, fill it up with the old data.
The goal of these systems is to distil this complex information into a more centralised knowledge management system that allows scientists to build on their own work effectively, and also their colleagues’ work, whether they are in the same lab or different labs across the world.
This is an edited version of a webinar held by Uncountable. It can be viewed at www.utech-polyurethane.com/events