Can you sense demand with social media?


While discussing ways and means to improve Sales and Operations Planning (S&OP) and forecasting, many a time business executives ask “What can we do with social media?" This was definitely NOT a usual topic in S&OP forum just a few years back! Most of the time, I push back the question with, “What makes you so concerned about social media in your usual month sales forecasting and planning?” There comes a pause… and then comes the usual answer, “Actually, since everybody is talking about social media… wondering... you know, if we can use that as well?"Demand Sensing with Social Media

Aha! Yes we can, but not because everybody is talking about social. Indeed, free flowing, unstructured textual data can be a tapped as a great source of information to understand its influence on future demand. The question is how do we do it? Allow me to elaborate!

Traditionally, an influencing customer’s sentiment; also expressed as WOM (word of mouth), market buzz etc., are regarded as good marketing tactics to improve sales besides the usual ‘pull’ and ‘push’ strategies. With the onset of social media, customer’s feedback, opinion, and expert view are in abundance, or at least more forthcoming, depending on the industry. Industries with high customer ‘involvement’ is likely to receive more social media postings (note postings can be negatively skewed most of the time). However, abundance has created another set of challenges i.e., the 3Vs of big data. I would imagine industries like automotive, hospitality, high-end electronics and consumer goods are likely to face the changeless of Volume, Velocity and Varity of social media data. However, I was surprised when a paints and coatings manufacturer approached us about leveraging social media data.

My recommendations would be to go beyond typical social media data, and think more broadly about the goldmine of semi-structured and unstructured data also available within the organization:

  1. Structured / Semi-structured data: Collected in certain intervals through third-party market research firms or online survey. These almost always include a customer satisfaction rating based on indicators like Top of Mind, intention of buy, likelihood to recommend and a whole set of brand perception or ad-recall data.
  2. Un-structured data / free flowing text : Typically from the following sources :
    1. Inbound mail : dealers’ complain on quality, customer complain on quality.
    2. In-bound and out-bound contact center log : dealers’ complain on quality, customer complain on quality – both on product quality and service quality.
    3. Social media: Facebook, twitter etc. where data is presently outside organization’s IT system.
    4. Trade forum / discussion group : People related to relevant profession, trade & customers express their opinion on own brand and on competition brand.

Technology, like SAS Text Analytics, has evolved to extract data from any of the above sources and further pull out the intelligence out of it. For example: social postings or inbound customer complaint e-mails can be scored to derive sentiment of that ‘text entity’. Sentiment are generally classified into ‘positive’, ‘neutral’ and ‘negative’.  Advanced text analytics techniques like the one offered by SAS, easily handles bad spelling, wrong syntax, para-phrase, diabolical expression through technologies like NLP, text parsing, Singular Value Decomposition (SVD), speech tagging etc. - and do this in more than 30 languages. An advanced illustration of leveraging textual data is the following SAS USER GROUP (SUGI) paper that shows how text mining can help one to understand demand and inventory situation in a hospital!

Since forecasting data is required in summary level, one can aggregate the sentiments for each brand or category. With sufficient history, it can be usually used as ‘influencer’ variable with regression based forecasting methods. However, other part of the forecasting hierarchy ‘location’ can’t be easily obtained in social media. Approximation rules are available to percolate the overall score into regional level.

Obviously we can take this thread of discussion beyond operational demand planning to more tactical topics like annual / half-yearly planning and forecasting with shorter history. I look forward to hearing your views and best practices!


About Author

Nilmadhab Mandal

Principal Systems Engineer

Nilmadhab is presently working with SAS and has been providing techno-functional consulting for SAS Supply Chain Analytics. He is an MBA and completed Advanced Program in Supply Chain Management. He has been invited to present at several seminars like International Business Forecasting Seminar (IBF) Amsterdam, Institute for Supply Management India and International Conference on Advanced Data Analysis, Business Analytics and Intelligence, IIM Ahmedabad (IIMA).

Comments are closed.

Back to Top