Top 5 reasons why automatic sentiment doesn’t bring love


Love is a rebel bird. That no one can tame”

I am amazed by the many initiatives around automatic measurement targeting  brands who desperately want to be loved.

Sentiment is very complex. Here are 5 good reasons, brand should forget about automation and invest in a culture and processes that enable their employees to engage with passion, build trust and develop their brands:

1- Sentiment is not linear.

A negative mention is usually easier to turn into a positive one than no mention at all.  The trilogy “positive, neutral, negative” doesn’t work. “No sentiment at all”  is surprisingly missing. Measurement has to be multi-dimensional !

2- Sentiment measurement is people-driven, not mentions-driven:

If I say 5 times that I am happy with my Mac, and after a while I say that it doesn’t work … I’m negative:  +5 – 1 = -1  🙂

Measuring by mention may be viewed as a good approximation. However, it does not enable to distinguish mentions from customers, non-customers, different segments, people using your product / services for a long time versus new buyers.  Should there be any “fatigue” or seasonality in the perception of your product  and the measure becomes incorrect and misleading

Some communities are more vocal than others. That’s the old Ideastorm problem. 95% of people who come to the Ideastorm forum talk about open source/ubuntu features… but it is only a fraction of the market. It just happens that windows users are a silent majority.

3- Negative for somebody is positive for someone else

Let’s take an exclusive brand. All negative comments like “it’s too expensive” actually reinforce the brand strategy and image. People who buy such exclusive products may well be very happy reading these negative comments about the price.

4- Influencer’s conversations are “special”.

A mention from a non-buyer in an “out of market” segment doesn’t have the same value as a mention from an influencer in your market .

5- IMHO automatic sentiment analysis does not work

Unless people invest massively in domain dictionary and training the algorithms.

If we step back, using automatic sentiment metrics is somehow treating the consumer/client as a three state robot with a simplistic positive/negative/neutral indicator.

I couldn’t find a better quote for this post than  Carmen.

L’amour est un oiseau rebelle. Que nul ne peut apprivoiser …” / ” Love is a rebel bird. That no one can tame”


One thought on “Top 5 reasons why automatic sentiment doesn’t bring love

  1. Dominique, Your points are ALL well taken and absolutely should act as a Caution Sign for putting extra-ordinary emphasis on “automated” sentiment analysis.

    That being said, whatever a company does to examine its reputation on the Web AND whatever it does to respond CAN have a positive influence on building a brand and its benefits for customers.

    So the only thing I would “disagree” with is the statement that “it doesn’t work.” I believe you simply have to accept the limitations and figure out how the collected data can be used successfully.

Leave a Reply

Your email address will not be published. Required fields are marked *