THE DIGITAL CONSUMER CONFIDENCE INDEX
To find out what consumers are thinking, you don't need to do an old-fashioned phone survey anymore. The Digital Consumer Confidence Index (DCCI)* is Networked Insights' weekly measure of consumer confidence in the US economy, drawn from the vast data set of conversation in social media.
Why Social Media?
The methodology for the traditional Consumer Confidence Index (CCI) is becoming increasingly limited compared to the power of social-media analytics. We've managed to find answers without asking direct questions — thereby removing biases. We add groundswell voices, naturally expressed, to the conversation about the health of our economy. It's no longer possible to get a complete picture of consumer attitudes while excluding this wealth of online consumer data.
Advantages of the DCCI
- Calculated on a weekly basis, faster than the traditional monthly report. Gives financial analysts and investors an essential advantage over those who are waiting for the monthly CCI.
- Weekly calculation helps to better identify trends. A monthly report with two down weeks and two up weeks will look neutral. The weekly reports can identify a two-week upward trend.
- Can better answer the question why the index has gone up or down. The model driving the DCCI can be used to drill down into consumer buzz that explains why confidence waxes and wanes.
- Avoids the increasing biases of traditional surveys. Pulls its data from millions of naturally occurring conversations across the web.
The Science Behind the DCCI
The DCCI is a consumer-confidence or sentiment classifier. The program reads a post and assigns a number to it representing the author's sentiment toward the economy. We trained this program on six months of social-media and Consumer Confidence Index data (May – October, 2009). That is, we trained it to find the signals in social media that correlate well with the Consumer Confidence Index.
The program is a combination of three areas of Artificial Intelligence: Natural Language Processing and two others (that you'll need to buy the company to learn). Prototype testing allowed us to estimate the program's error at +/- 10%. Each monthly DCCI calculation has been well within this estimated error so far.
*The DCCI is managed entirely separately from the CCI and has no relation to the Conference Board.