The private healthcare company Bupa is the middle of a much needed business intelligence (BI) transformation, centralising data into a single data warehouse and simplifying reporting to deliver insights at far greater speed.
It's a tried and tested formula for Julian Pimm-Smith, director of data and information services at Bupa, and his team from their days at Pret a Manger: layering SAP's Business Objects enterprise on top of an SQL server data warehouse, with the Business Objects Universes semantic layer in between.
Speaking to Computerworld UK during the UK and Ireland SAP User group conference in Birmingham last week, Josh Morrin, senior reporting developer at Bupa, talked us through the project after he followed Pimm-Smith from Pret in July last year.
Seconds rather than hours
The first step was to centralise data and reporting for the organisation, instead of the old model where "analytics happened disjointedly around the business, using a lot of different tools," Morrin said.
"They didn't really have a data warehouse, they referred to it as a data warehouse," he said of the old model.
"Essentially it was data marts that were just replications of the source tables, mildly manipulated. None of it was dimensionally modelled, so now we are building it for reporting purposes from scratch and layering the Business Objects Universes on top."
Where analysts previously were making their calculations "in the report level with lots of clever coding", now with Business Objects reporting can be more consistent and, crucially, much faster.
"There are still reports that people run for 36 hours," Morrin said. "There was a table we imported into Business Objects that was 440 million rows and counting.
"If you're trying to reindex or filter or aggregate that down at report level you have no chance, so the fact our data warehouse can swallow that is new to these people."
Now it is a case of Morrin and his team bringing these analysts into this new world.
He said: "It has changed what we do because we are essentially spending a lot of time selling our new capabilities to little MI (management information) teams, because we are building a big platform of Business Objects for the 500 or so analysts at Bupa to try and get them all using Business Objects instead of whatever else they are doing."
The advantage of this new centralised capability is primarily speed, as the old model involved filling out four-page paper forms to request certain data items and reports.
"The old way the team works is they send you a four page form where you would ask for your data items, what systems are being affected, all of this stuff that's like - why would they know that?" Morrin asks. "So we ditched all of that and it has kind of opened the floodgates."
This means shifting to a more service-oriented outlook from Morrin's team, as well as providing some self-serve capabilities for analysts.
"Now people are generally able to just have a chat, by email or Skype, about what they want and we can work backwards from there," he said. "So I would say service delivery has been our best thing because it is much quicker."
The self-serve capabilities for analysts involves a set of "input controls to change which group or claims base or product they want to look at" and quickly run a report.
Now, instead of running Cognos on top of some loosely defined replication tables, running Business Objects on top of a central data warehouse allows Morrin and his team to run reports in seconds not hours.
"We would expect to have that information out on the same business day, rather than before where you would be talking about two weeks, as there was a whole other team to test the reports," he added.
Naturally with a change like this there are teething issues, especially around change management and stubborn analysts.
"They don't all use Business Objects to report off currently, and realistically I can't see all of these teams using it," he admitted.
"Transactional reporting and things that can be scheduled and automated is fine, but some of these teams are doing predictive analytics, data science and data modelling and Business Objects doesn't quite do that."
This means Morrin and his team run a separate enterprise R platform, supported with Python installs.
"We still have some analysts who have their Tableau and PowerBI stuff on top of the data warehouse," he said. "Those users are the ones we are trying to phase onto Business Objects. Those using SAS and R are different."
"I would say we are 12 months off from having all of the reporting systems properly in the data warehouse," Morrin said, meaning the organisation still has plenty of work to do. "Lots of the main pertinent ones like finance and insurance are in there, now we still have to get them looking at it."
This is particularly important when considering the GDPR and audit responsibilities the company has to maintain - "especially with the data we have, very personal data," according to Morrin.
The key here is to start automating some of this reporting for audit. "We are always OK and never get a bad result from an audit, it's just the time it takes to maintain that," he added.
To combat this Bupa is using a set of tools from specialist Business Objects bolt-on vendor 360 Suite, which is "designed for that, to answer all of the questions you typically get from the audits around who can touch what or where items are being used and who has access to what".
"So the security guys like Business Objects because you can identify groups and just cut that branch off in terms of access, or if someone has run something you can identify that as it all links back" Morrin said.