The right questions
Written by Alison Ebbage
Retailers seeking quick ROI and robust results are looking into in-depth data analytics. The idea is that by mining existing data in a new way, behavioural traits or procedural abnormalities can be identified and flagged up for further investigation, addressing both shrinkage and fraud.
Richard Dorf, managing director at Pxtech, comments: “The challenge for any retailer, and especially senior management who are not actually on the shopfloor, is how to get better visibility of potential issues and then how to draw evidence once an issue has been positively identified.”
Indeed, retailers dealing with high volumes of transactions find it impossible to keep track of each and every event that could, potentially, be suspicious. The beauty of data mining is that it can be set to look for anything that deviates from normal parameters. Each store can set its own parameters. In addition the software does not require any additional data sets to be gathered, rather it makes best use of existing data by bringing it out of its original silo and analysing it in a single view format. This shows up patterns of behaviour which might not otherwise be identifiable if the data stayed in its original ‘home’. It can then be investigated.
Richard Paterson, business development manager at HiCom, says: “Retailers deal with large volumes of mostly low-value transactions, making it hard to spot patterns or individual incidences of theft. It is hard to prevent determined people from stealing but with the right software you can see where the weaknesses lie or see where a pattern is emerging. Employees knowing that you have the ability to monitor also acts as a deterrent.”
Getting it right
Indeed, asking the right questions of existing data can be pretty illuminating. The key is to programme the software with open-ended questions and be flexible enough to change those questions as things move on. In an automated sense this is an exemption-based system knowing that it needs to look for x and y and perhaps issuing a regular traffic light alert report. This will identify issues and also help to see if the parameters need changing slightly to get a more granular view.
David Snocken, commercial director at IDM Software, comments: “Procedural weaknesses that might lead to shrinkage or loss could be something like a time limit on refunds not being adhered to, a cash refund level set temptingly high or an alarming level of price overrides being allowed by the PoS terminal. By looking at high instances of this sort of thing, the management can get an idea of what is really happening on the shopfloor without needing to look at each and every transaction.”
By way of example one retailer was seeing shrinkage levels on leaks in terms of the difference bought and sold by weight. Further investigation revealed that the leaks were not being trimmed enough and, although the retailer was paying for the full weight, the scales were not physically big enough to register that, and only charging for the portion of the leak that it could weigh, hence the high shrinkage levels. Clearly, then, having the tools to provide decision makers with an actionable report, where data has already been deciphered, is really useful.
But data mining is not just on a piece by piece basis. It can also be used to compare different stores and feedback useful reporting on what might be going wrong. For example, says Paterson, it can pull in footfall, store layout, location etc and look for correlations to identify patterns. “Two stores that are, on the face on it, very similar might have different levels of theft. Is the one with less theft accurately recording all incidents? All the data stores combined can help to identify what makes a risky store and where things are going wrong,” he says.
The software can also link in to other systems or fraud detection methods such as tags, (electronic article surveillance), or the gates at the door. One of the most crucial linkages is to CCTV so that, when something that is already on a watch list happens, the manager can use the CCTV to see the actual events.
For example, a common fraud where there are large volumes of cash is a disproportionate amount of cash drops where the cash drop is being used as an excuse to open the till, or price overrides where an item is registered as 0, or less than its worth, and the employee pockets the money. Clearly this is valuable in its own right but being able to, and being seen to have this functionality also serves as a strong deterrent to employees.
The combination of the software and the CCTV can also be useful when tackling customer fraud and the need to share that information and follow the right procedures with third parties such as the police of local crime reduction schemes. Again the key advantage is that by taking the data out of its silo and having a single point of contact, the reporting is done once then passed on to the relevant people, be that HR, the police, head office, an external security company or other third parties.
Paterson explains: “Once the information has been recorded it can be pushed out to the right people automatically with minimal repetition. As the report is held centrally it can also be updated and so if one user updates the report, all the other users will automatically be able to see the latest version without specifically needing to be informed of an update.”
But perhaps the major selling point of many of the new breed of data mining solutions is that they come as a hosted service and this means that they are cheaper, ROI is swifter and the system goes down as an operational cost rather than a capital expenditure. This makes it more readily available to smaller players who had previously been unable to get the right balance between loss mitigation and IT spend.
Dorf comments: “There has been a significant upswing recently as retailers realise they need much better time efficiency on this sort of thing. And as prices come down the cost versus savings argument becomes more compelling as well.