Is Bigger Data Really Better Data?
By Michael Schubach, Strategic Deployments and Program Management Director at Infor
The Digital Age is truly the era of knowing many things, but given our access to incessant tribal chatter across social platforms, the equal coexistence of facts and alternate facts along side news and fake news, one could rightly feel as though we humans may have access to more information than is necessary or even helpful. Despite such misgivings, I think we are better people because of our access to plentiful data. Nonetheless, I feel compelled to revive an analogy that was first offered to me when I was a Cub Scout: Data, particularly 'big data,' is like fire - used wisely it can light our way, allay our fears and give us warmth. Used foolishly, it can be destructive and devastating. The most useful objective, therefore, is not fire prevention but meaningful fire management - learning to discern the difference between quantity and quality, and react accordingly.
But all data is not created equally. As I look across the kinds of information that we use to find and serve our guests, I see four distinct data types, distinguished by their method of manufacture:
When we specifically elicit information from a guest or customer, this is active data collection. We can either ask in person (What is your last name? Could you spell that for me, please?) or on a form, but either way we are telling the provider what we wish to record so that we may use that information in a manner that somehow benefits him or her - at least, that's the theory.
Active data is typically the best information you can retain; unless your respondent has something to hide or simply does not want you to have access to the requested information, what you get is accurate according to the best source possible. It's important to note, however, that even active data has a shelf life: last name, address, age (as distinct from birthdate), marital status, smoking preference, need for ADA accommodation or assistance can all come and go. "Eternally true" is not a property that we associate with guest data; there is a need to ask and reconfirm all kinds of data on a regular basis.
During the course of conducting routine business within a hotel, a great deal of routine data is generated. Keeping this data available and using it to make future operational decisions is an example of passive data collection. One might assume that since the data is factual (e.g., the guest slept in a king room the last time s/he stayed at the hotel), it is a realistic reflection of preference (the guest prefers sleeping in a king-sized bed). It is also reasonable to assume that there was some deliberate data collection step involved in the assignment of the room (Do you prefer a king bed or two double beds?). Thus the king room, duly noted, would seem a clear preference for future room assignments, correct? The official answer in this example is "maybe or maybe not." Reusing this type of data implies that an active choice was made and a clear preference is reflected in its presence.
What is not clear is if the prior circumstances are an expression of an ongoing preference. Even less clear is what was actually being selected at the time - what if the question the guest answered was "Do you prefer two double beds on the first floor facing the freeway or a king bed on the concierge floor overlooking the waterfalls?" Did the guest select the king bed or acquiesce to the situation? What was actually selected? Was it a king bed or a higher floor or a premium view? Asking the unfettered question, "All things being equal, what kind of bed do you prefer?" is entirely different than giving your guest a choice between room 104 and 2227. Passive data collection produces true facts that can be misleading when used to make derivative decisions.
From my perspective, assumptive data is that which is based not on me or my personal responses, but rather because I "fit the profile" of a particular affinity group about which data is known. Assumptive data exploits the old marketing maxim that calls out flocking tendencies of "birds of a feather." An example from outside the hospitality industry that illustrates the risk of using assumptive data is the fact that I've lived in a "deep red" southern state for nearly a decade. This could imply something about my cultural values and/or political leanings - or it might not. In an industry example, a revenue manager from a very large chain that focuses on moderate priced hotels for the business traveler, once told me that if I call on a Monday or a Tuesday to request one night's accommodation within that same business week then, with statistical certainty, I will request and will be willing to pay extra for a king-bedded room.
Statistically, I'm sure she was correct. I, however, always get a room with two double beds so that I can throw luggage, clothes and papers on the other bed. Assumptive data works well if your guest is not a statistical outlier; for outliers your data point, regardless of probability, is wrong.
Another way to get actionable data is to collect and retain large quantities of raw data, assuming that the more you have or observe, the more apt you are to extrapolate meaningful information from it. This is essentially the "secret sauce" of big data: the whole is more revealing than any of its parts. Parsing a passel of data is how we compute averages, spot trends and evaluate statistical possibilities. Armed with this information we may not always win, but we are certainly playing the odds and betting with the house; this is "mathe-magic" of aggregated data.
Consider my earlier question of the king size bed with the waterfall view. In that one single transaction it's difficult to determine the guest's motivation in the selection. If however, a guest stays eighteen times a year, and fourteen of those times are in a king room, we can, with gathering confidence (nearly 90%), be justified in thinking that the guest's room preference is being accommodated. The best evidence to support our conclusion is the frequency with which the guest rebooks the hotel; something must be to the guest's liking, even if it's just location, location, location. The lesson we take from this particular example is that more data can indeed make us righter.
Another example of the power of big data is the way it can improve hotel service. Because today's systems have the capacity to process and store massive quantities of data, service rendition can be analyzed in detail and can be used to predict service outcomes. Properly architected service optimization systems produce a wealth of data as they monitor and record request fulfillment. By tracking lapsed time of each incremental step from initial order through final fulfillment, we can track staff productivity and harness the data for predictive purposes.
For many years, the industry used standard statistics against which performance was measured; for standard guest rooms (think about a roadside Holiday Inn here), a housekeeping room attendant was expected to turn the room in thirty minutes, finishing sixteen rooms in an eight-hour shift. Of course, a slew (a precise term for an imprecise quantity) of variables affected the actual outcome. Sixteen rooms was simply a target number that was deemed to be challenging but practically attainable, and variations either "came out in the wash" or were used to identify consistent over or underachievement. It was really as much practical measurement as circumstances permitted; short of holding a stopwatch to a room attendant, it was a rule of thumb that generally worked.
But today, the memory stopwatch is abundantly available, and can be used for every room attendant on every room assignment. And those slew of variables? We can passively record almost all of them: we can know the size of the room assigned, the number of occupants lodged within, the level of service requested or due today, the time of the assignment (for shift or day-part analysis), the travel time from room to room, the inspection time, any come-back requirements to correct inspection discrepancies, and last but certainly not least, the room attendant him or herself, with whatever tracking is permitted or useful there, things such as age and years of experience. We know everything but the weather conditions.
Now, let all these data points collide over a span a thousand room turns - that's 12.5 full-time weeks by the old reckoning - and a desk clerk can be told that the room s/he needs, a king-bedded junior suite with terrace and a park view, that was vacated three hours ago, requires full checkout service that takes the assigned room attendant forty-four minutes and forty-five seconds to complete. It will take the inspector an additional twelve minutes and fifteen seconds to arrive, inspect and release the room, with a nine percent chance of requesting corrective action, which could (but likely won't) add another fourteen minutes to the completion. Since the room is already in progress, the attendant having started on it thirty-nine minutes ago, the room is now estimated for completion in exactly eighteen minutes. Sending both the attendant and inspector an automatic "guest is waiting" request to expedite the room will shave six minutes off that estimate. That leaves the guest a theoretical wait time of twelve minutes - just enough time to enjoy a complimentary cup of coffee while the room is finalized.
So does data really light our way, allay our fears and give us warmth? Maybe in a somewhat strained analogy, one out of three isn't bad. Data is most certainly a part of enlightenment. While it might be short on warmth and reassurance, it is what makes guest recognition and superior service a practical reality. It allows you to personalize guest service, assuming, of course, that you train and field personable, hospitable service providers. Understanding data collection, interpreting it well and making it accessible when appropriate is the key to surpassing guest expectations and delivering memorable and differentiating hospitality experiences.
Infor is a global leader in business cloud software specialized by industry. With 17,300 employees and over 68,000 customers in more than 170 countries, Infor software is designed for progress. To learn more, please visit www.infor.com.
Infor customers include:
- 19 of the top 20 aerospace companies
- 9 of the top 10 high tech companies
- 18 of the 25 largest U.S. healthcare delivery networks
- 18 of the 20 largest U.S. cities
- 19 of the top 20 automotive suppliers
- 17 of the top 20 industrial distributors
- 15 of the top 20 global retailers
- 4 of the top 5 brewers
- 17 of the top 20 global banks
- 9 of the 10 largest global hotel brands
- 8 of the top 10 global luxury brands