3 years with Sysomos MAP - what I've learned

Limited API access to some networks. Though this is hardly unique to Sysomos, the limited APIs provided by closed-networks like LinkedIn and Facebook, and others like Instagram, often disappoint our users, who may have unrealistic expectations as to what we can report on. In reality, Sysomos MAP is highly focused on Twitter, which is often not our core channel. Though Sysomos is bound by the limitations of the APIs provided, perhaps some creative innovation would enable limited but interesting observations. Identifying LinkedIn statuses that were simultaneously pushed to Twitter, for example, gives a window into that otherwise private world. Scraping of LinkedIn's trending topics might be revealing.

Sentiment analysis is automated and therefore only moderately accurate. Some advanced customization might enable us to get better results, perhaps by uploading custom dictionaries of positive, negative or neutral terms. For example to a loss-adjuster or insurer, many terms like flood, bankruptcy, damage, fraud risk etc may in fact be used positively as part of a marketing proposition or work undertaken ("X company insurance saved us from awful disaster"). Automated systems will always struggle with these. I wish I could make mass-edits to such inbound comments that would learn and persist.