Radio frequency identification (RFID) technology presents businesses with a vast range of opportunities, and some meaty challenges to boot.
Many firms have been piloting and prototyping RFID over the past few years, including high-profile companies such as Wal-Mart, Gillette, Tesco and Marks & Spencer, with much of the focus on the physics of getting readers to read data from tags in an unfriendly physical environment, such as a warehouse.
But once the hardware infrastructure is in place, the key challenge is to store, transport and integrate the captured data with corporate software systems. This is where the challenge for IT comes in: the management of data for RFID.
There are three places where RFID data needs to be managed. The most specific to RFID is at the edge – the raw RFID data that is recorded by readers and its translation into business events. Edge data comes in huge volumes; the number of business events is less so, depending on the level of events to be captured. For example, asset tracking could describe moves between locations as a business event, but RFID readers would read the position of the item many times a second.
In common with data from any application, RFID business event data needs to be transformed and aggregated into meaningful information, locally for departmental applications such as stock positions across a warehouse, and centrally for a company-wide view.
Business outcomes must be tied to the technology programme, for RFID no less than any other project. The more that is known about the required business outcome, the better RFID, its IT components and integration can facilitate this.
Business managers should be responsible for defining the business events, what combination of reads makes a business event, and what information from a read should be passed through to the event.
The business also needs to define the amount of data and level of aggregation required for analysis and reporting purposes. The more data that is aggregated centrally, the more care needs to be taken when sizing for data volumes.
Local analysis requirements, such as warehouse performance, may need additional software and staff to view reports and take action. Performance management tools can help, and integration with business process management systems can automate some actions.
The analysis requirements should be considered in line with existing central business intelligence (BI) systems to make sure data from RFID sources can be integrated with applications. This sounds obvious, but the temptation to consider implementing RFID in isolation should be resisted.
Not only should these events and aggregations be defined, but changing them should also be possible.
As BI practitioners are well aware, users do not know what they want to analyse until they see something that is not what they want. Aggregating more, less or different data should be possible within preset limits.
Once RFID is integrated with other applications, data quality rears its ugly head. Data quality is key for firms even without RFID, and raises a number of issues.
The first hurdle is the difficulty of quantifying the problem of data quality, combined with a reluctance to spend time and money finding out bad news.
At best, poor data quality leads to inefficiencies, as users have to match and amalgamate incorrect information manually, and at worst it leads to accounting irregularities when revenues are incorrectly booked – which these days can mean prison sentences.
To successfully address data quality, an organisation needs to do a number of things: understanding the high cost of low-quality data is vital, followed by the development of a data profiling and clean-up plan.
Once the existing data is as clean as possible or necessary, the business needs to take ownership of key data items, to manage and monitor data quality going forward.
Alys Woodward is a senior analyst at Ovum. She will be speaking on this topic at the Softworld Supply Chain & Logistics event on 29-30 March 2006 at The Pavilion, NEC, Birmingham.
reader comments