Adding the Analytic Dimension to RFID
Earlier this decade, Wal-Mart first required its top 100 suppliers to put radio frequency identification tags on shipping crates and pallets, then extended the mandate to include all suppliers. Thanks to the Wal-Mart mandate, the first wave of serious enterprise RFID deployments are on the books, and the result is a C, maybe a C-, in terms of ROI and maximum utilization of the technology. Many have completed their Wal-Mart compliance requirements so they can continue to do business with the retailing giant. However, few have taken the opportunity to go beyond, into the other areas of their business. Fewer still have incorporated other data sources into their application or produced analytic value with RFID data.
This is not to say that companies outside Wal-Mart mandate have not been adopting RFID. Retailers utilize RFID in their stores to manage temperatures, control theft and shrinkage, influence store design and product placement and even experiment with self-service checkout. Indeed, with billions of tags ordered, there are numerous uses for this tiny technology. These devices track cows, passports, toll tags, credit cards, clothing, concert wristbands, pharmaceutical pedigree, laptops, ski gates, neighborhood gates, shopping carts, pets, robotic companions and humans. From a Wal-Mart perspective, it’s the pallets of goods sent to its distribution centers.
One of the setbacks has been with security. In particular, the passport utilization of RFID has come into question. Post-rollout concerns about hacking has led our government to recommend they be protected in cover sleeves. The other major obstacle stalling progress is the economy. However, as an economic foundation seems to be settling under us, companies are becoming more amenable to calculated projects that have demonstrable ROI. RFID is a technology, in many cases already deployment somewhere in enterprises, that offers many positive returns.
There are passive, active and in-between tags. Passive tags have no internal source, are the smallest tags and have an indeterminate but long lifespan. Passive tag antennae “wake up” when hit by an incoming radio frequency, which represents a transaction. A corresponding transaction, should you want it, can occur when the tag leaves the reader’s range. Read range on passive tags is up to about 10 feet. Contrast that with active tags, which have a read range of several hundred feet. These larger tags have their own power source. If applicable, the reader can be a writer as well and can store information on the tag to be read by the next reader. The batteries in readers have about a five-year lifespan.
By combining the tangible item, exact location, timestamp and other potentially factors such as temperature with a master list of properties, which fully characterizes that item, the meaningful information a company can utilize is generated. Through well-placed (usually fixed) readers, the company knows the status of its interesting items at any point in time. RFID readers can penetrate without a line-of-sight and have tremendous abilities to read tags in rapid succession. This entire process is aided and abetted by the Electronic Product Code (EPC), which is playing the role of global item master.
Once the data is being generated, taking appropriate actions, both operational as well as analytical, are the key to success and ROI. Operational decisions are reactive and usually involve limited data – mainly that which can be brought into the picture immediately in support of an operational misconduct that needs remediating. Examples are when the temperature for the milk falls below threshold or if a product moves through the exit without having just previously moved through checkout.
For example, what good is an RFID tag read if you cannot ascertain what the product attributes are (intelligence associated with the EPC), what location the GPS coordinates represent and, if a customer is involved, what his/her normal or preferred pattern of behavior is? It is also helpful to compare the item and timestamp with its normal or preferred locations. What about the prior locations of the item/person? Should the last five minutes or 10 reads be cached and utilized in operational decisions?
The real promise of RFID lies in collecting, managing and analyzing massive amounts of information. The information is not only in the streaming reads, but also in the previously streamed reads, which are stored and provide context to the current reads as well as analytical value. This is the contribution of the data warehouse, where the historical data is collected. To assist in the operational actions, wisdom is derived from the data warehouse and fed back into the operational environment. An operational, interactive data warehouse is essential to RFID analytic success.
The other essential information management component is a widely applicable master data management store, which is based on the EPC. The EPC code, maintained as an industry standard by a third-party consortium, has 96 bits, which allow it to uniquely identify every physical asset on the planet – for years to come! Currently, the focus is on consumer products that are individually tagged.
For now, there are several opportunities to extend RFID capabilities within organizations. These consist of building an organization’s operational reactivity and analytic capabilities with an information management architecture supporting RFID data integrated with the rest of the organization’s data.
Smarter Inference in Receiving
Readers frame the entryways to container ship drop-off points as well as store distribution centers. Reads, assuming accurate tagging of products overseas, track what has been delivered. Once trust, is established with the supplier, through this verification, the retailer can get a very accurate picture of the products delivered over time, without slowing down the supply chain at all. Currently, such full lifecycle views are possible in only limited fashion.
Stronger inference techniques also could be part of the RFID chain. The read process is not foolproof. If 90 percent of products are certified in the read process, should the receiver assume the shipment is 100 percent accurate? If so, the implication here is that not only do we impute reads into the operations, but also into the data warehouse, reflecting the reality that the company is operating under. A certifiable process of imputing can tide a business over until the tag read process is perfected.
Retailing to the Customer of One
If you’re in retail, making the next best offer to that customer is imperative. However, in most retail situations, total knowledge of the customer is not natively known by the employees doing the interaction. The corner drug store is a relic of the past.
Several customer touchpoints can be considered in optimizing the next-best offer. Partially, these include:
- Operational factors, such as the customer’s current movements throughout the store; the products the customer has in his cart; the products the customer triaged or otherwise spent time near, but did not put in his cart, what the customer is wearing; and what credit cards the customer is carrying
- Syndicated RFID data could include when and where his wardrobe was purchased, and the customer’s movements in previous visits to the store.
- Traditional syndicated marketplace data, such as credit history and FICO score, customer demographics, customer psychographics, and customer econometrics.
- Analytical factors, including the customer’s previous purchases and customer lifetime value.
- External elements like news and weather.
Store entry is the best time to identify the customer. Otherwise, all of these touchpoints are unknown until checkout, where they are about 10 percent as effective as when the shopping is actually occurring.
Just dumping this information onto the sales agent is going to be mind-boggling. I would put the information from the traditional syndicated marketplace into the analytic data warehouse layer, except that it is quite possible the customer is unknown to the retailer, in which case the retailer would not have sourced that information. Of course, the sales agent would not have the customer’s previous purchases or lifetime value to bear on the transaction either.
Varying levels of maturity will arise in the retail experience that will comprise not only the absorption of touchpoints such as these, but also, and just as importantly, synthesis of that information, yielding the proper action to take with the customer. Extending channel connectivity in the store experience to outlets such as cart display, cell phone texts, aisle display and human contact would be another factor.
Some of the most interesting forays into RFID have come outside of traditional product tagging and are in the area of pharmaceutical pedigree. The use of the inference techniques is unavoidable in any application, but it is especially important in high-value products such as pharmaceuticals. Each drug container may require an electronic pedigree with the aforementioned write activity at each point in the manufacture and distribution of the drug. With the high-risk downside of counterfeits, smarter inference is required in these processes.
The pharmaceutical company’s ability to apply analytics to put the data into the perspective of acceptable movements will provide even greater value to the tracking currently being done.
The use of RFID could also be extended into the consumption end to ensure not only that the right drug from the right pedigree got into the right consumer’s hands, but also that it is properly consumed. RFID readers issued to homes could at least ensure the passing of the container through a reader at the container’s destination, which would imply triage and presumably consumption. Communication could be issued through customer channels not unlike the store example earlier.
The advantages of RFID are unmistakable, and early adopters have realized ROI. The gating factor to more projects and benefits seems to be the ability to incorporate analytical data into the processes. A few industry leaders are already reaping the benefits that accrue exponentially once analytic decisions begin to be made with the empirical basis of RFID tag reads.