Last week I finally was able to get my 1 year old daughter vaccinated against the H1N1 flu virus. There have been widespread availability issues with the vaccine throughout the US. When a batch of supply is delivered there is a mad frenzy of parents who line up outside the medical facility as if they were waiting to buy concert tickets. I recall showing up to the pediatrician’s office at 7AM on Halloween Saturday this year only to discover that I was the tenth person in line. Although, I did get my daughter vaccinated that day I found out that her age group requires a follow-on booster shot 4 weeks later. And every time we have tried to get the follow up shot she has either been sick or the vaccine has been out-of-stocks. Stick with me – I will explain how this does relate to both Google and the Supply Chain.
Fortunately, the H1N1 virus pandemic has not been as serious as many expected last spring. There have been over 10,000 confirmed deaths worldwide from the subset of the population which has become seriously ill. However, most people only experience mild symptoms such as fever, sore throat, cough, headache and muscle pains. Massive efforts have been undertaken by public health authorities to track the outbreak of the H1N1 flu strain. Both the European Influenza Surveillance Scheme (EISS) and the US Centers for Disease Control (CDC) have been closely monitoring clinical data from health care providers to quantify the pandemic.
Tracking Flu Trends on Google
Perhaps, the best source of data on flu outbreaks is Google. The search engine leader launched an application called Google Flu Trends in November 2008. Officially, Flu Trends is called a “syndromic surveillance system,” which is a form of “infoveillance.” What it does is to analyze end-user search patterns in an attempt to map flu outbreaks geographically. Google states:
“We have found a close relationship between how many people search for flu-related topics and how many people actually have flu symptoms. Of course, not every person who searches for "flu" is actually sick, but a pattern emerges when all the flu-related search queries are added together.”
Comparison of Google Flu Trends data to US Public Health Agencies (Source: Google.org)
Flu-related topics might include symptoms such as fever, fatigue and headache. The search terms could also include products such as cold medicine, tissues and Orange Juice. Google has found a very high correlation between its flu models and the actual statistics reported by the European and US Centers for Disease Control. There is one key difference. Google’s data is several days, if not weeks, ahead of the public health authorities. The difference is no surprise to me. Google is analyzing search patterns in real time, while the public agencies are analyzing historical data from clinical providers. The situation is very similar to the retail supply chain. Manufacturers of consumer products all rely upon getting copies of Point-Of-Sale (POS) transactions from their retail customers to analyze demand patterns. Using POS data is an improvement of their demand forecasting techniques of the 1980s and 1990s which were relatively uncorrelated with consumer activity. However, much like the CDC data used to analyze flu patterns, POS data is historical in nature. And historical consumer activity is not necessarily a good predictor of future behavior.
Beyond Philanthropy – Commercial Applications of Google Flu Trends
Google Flu Trends is a free service from Google.org (not dot com), which is a philanthropic arm of the technology innovator. Google.org has altruistic motivations behind the various applications that it creates to better society. But I have less noble ambitions for these applications in mind. Instead of limiting search engine data about cough medicines, tissues and Orange Juice to just public health applications, why not extend the analysis to commercial applications. For example, I suspect that the manufacturers of these cold-related products would substantially benefit from using Google’s search engine data in the demand forecasting process. Google has a wealth of information about consumer purchasing plans in its search engine data. The best part about these demand signals is that they are indicative of future consumer behaviors rather than what occurred in the past.
Search engine data can be quite revealing when analyzed. Remember the case of AOL user 4417749. In 2006, AOL published copies of stored search queries from 650,000 of its users to general public. The data was de-identified removing screen names to make it anonymous. However, through relatively straightforward analysis, NY Times reporters were able to uniquely identify the people conducting the searches. User 4417749 was determined to be Thelma Arnold, a 62-year-old widow who lives in Lilburn, Ga. My point is not that Google should violate the privacy of its users, but rather that relatively simple web log and search engine data can provide very powerful insights into consumer behavior.
Google Insights – The First Step
Google is already moving towards this type of model with its Insights product Beta. Google Insights allows you to compare search volume patterns across specific regions, categories, time frames and properties. For example, you could inquire how many people have searched on the term “iPhone” in the Baltimore, MD area within the past 90 days. By comparing historical search patterns you could have an indication of whether demand is rising or declining for Apple’s phone within that geographic region. Such Insights could be added into demand forecasting models along with other inputs to assess what stocking levels should be in the AT&T retail locations which sell the iPhone. The same process could be applied to home appliances, automobiles, consumer electronics and a wide variety of other merchandise categories. Google could sell syndicated, zip-code level data to manufacturing companies much as Nielsen or IRI do with aggregated POS transactions today.
Today, Google Insights is a free service today with limited flexibility in search parameters and insufficient granularity to provide neighborhood level demand patterns. However, the limitations are not a function of the lack of data, but rather the fact that Google has only begun to develop the Insights service. In the coming years, I think companies which can harness the power of user-generated demand data from the Internet will have a competitive advantage of their peer group. Read more about this concept in my posts entitled Beyond POS – Looking Forward not Backwards for Demand Forecasting and Why Amazon.com has the Best Demand Forecasting Data or my white paper – Download Beyond Point-of-Sale – Using Web 2.0 Technologies for Demand Planning.
