Member Avatar for InsightsDigital

We know that when school is in session during the Fall, it can also be a season of communal flu and the cooties. Thus, what do you think of Google's FluTrend - http://www.google.org/flutrends/ ? It is good for writing a blog posting or a report for a kid, but from a bigger perspective, do you think it can help nations prevent flu by looking at the trends?

  • 2 Contributors
  • 1 Reply
  • 215 Views
  • 7 Hours Discussion Span
  • Latest Post Latest Post by MktgRob

Dani AI

Generated

A short, practical addendum to the thread: ’s prompt and ’s point about extrapolation hit the key issues. Google’s Flu Trends used correlations between aggregated search queries and CDC influenza‑like‑illness (ILI) reports and, after its public launch in 2008, produced near–real‑time “nowcasts” (often 1–2 weeks ahead of official reports). (nature.com)

Independent evaluation later showed systematic problems: model drift, media‑driven spikes in health searches, and opacity about which queries and weights were used. A 2014 Science analysis documented major overestimates (for example, large errors in 100 of 108 weeks around 2011–2013) and framed the failure as a lesson in "big‑data hubris." Google stopped publishing live Flu Trends data in 2015 and moved to share the underlying signals with public‑health researchers. (gking.harvard.edu)

What that means for national planning today: search‑based signals can be a useful early‑warning input but should never be the sole basis for policy. Practical checklist:

  • Treat search trends as an early indicator to trigger investigation (not as a case count). (gking.harvard.edu)
  • Validate and recalibrate models continuously against sentinel, lab, and EHR data (quantify bias and error margins). (gking.harvard.edu)
  • Combine multiple streams (search, social media, participatory reporting, clinical/EHR, lab confirmations) with ensemble methods — ensembles outperform single‑source models. (pure.johnshopkins.edu)

In short: the idea behind Flu Trends was sound and useful for nowcasting, but its failures show why nations need transparent models, routine recalibration, ground‑truth validation, and multi‑source ensembles before using digital signals to guide prevention or resource decisions. (nature.com)

Member Avatar for MktgRob

Having worked in market research I have an intimate understanding of how they get to their final numbers with the inevitable margin of error. The term is called extrapolation and it basically means they take a representative sample of a larger population and expand out the test results from the sample to correlate to the larger population. The margin of error allows the reporting agency to account for the unpredictable.

The problem is that unless it is an extremely scientific study done by a neutral agency, the numbers are not always that reliable. In the case of the Google FluTrend, it may help or it may not. I think that this more of a case of Google taking a hot topic and ensuring traffic to their sites by providing additional information to an information driven society.

Be a part of the DaniWeb community

We're a friendly, industry-focused community of developers, IT pros, digital marketers, and technology enthusiasts meeting, networking, learning, and sharing knowledge.