It is well accepted that historic accounts are of limited value when it comes to forecasting future performance. Results for the last period merely record what has happened and in the context of the highly volatile and uncertain markets (which many companies have to endure at the moment) provide little insight into what may happen in the next period. Gary Simon, FSN managing editor considers the potential use of social media KPI’s in business forecasts.
Traditional measures such as profits ‘year-to-date’ and revenues ‘year-to-date’ (so called lagging indicators) provide some insight, added to which variance analysis against equivalent trading periods in the past can provide some illuminating information about general direction of travel and possible reasons for under or over-performance. But when it comes to looking six months or even a year ahead the dependability of financial indicators becomes less reliable.
So conventional wisdom encourages businesses to examine non-financial (forward looking) key performance indicators in order to develop more accurate forecasts. Generally, organisations are quite bad at developing and measuring non-financial indicators with few actually making it onto dashboards and board commentaries. It seems that management feel more comfortable with actual numbers than non-financial indicators that are often considered too ‘touchy feely’ to be taken seriously.
One of the reasons for this is that non-financial indicators are often measuring sentiment, for example, customer and employee satisfaction and, sentiment by definition, is an ethereal concept that does not lend itself to quantification. Or is that a misconception?
Take for example ‘Consumer confidence’ which weighs heavily in news reports about the state of the economy. A variety of ‘confidence’ indices are thought to provide a good indication of buying intentions – especially when movements are measured quarter upon quarter. Although these indices are well established on a macro-economic level, individual companies are less inclined to use them in their own setting.
Yet sentiment can prove extremely helpful. A high degree of customer dissatisfaction can be an accurate harbinger of a drop in sales revenue, a rise in complaints, credit notes, unpaid invoices and returned goods. Similarly a decline in employee satisfaction can easily translate into poor productivity, high levels of absenteeism and an increase in attrition and recruitment fees.
Of course capturing satisfaction levels of whatever hue involves a lot more work than harvesting transactions in the general ledger. Companies have to work hard on customer interviews, employee exit interviews, surveys and assessments to glean reliable data. This expression of sentiment then needs to be quantified and calibrated with financial performance in order to derive useful insights about future performance.
But Social Media presents (potentially) a new rich seam of information about sentiment. Facebook, Twitter, Blogs and discussion forums are littered with unsolicited comments about companies, their directors, employees and products. Reading these comments provides an indication of sentiment and indeed, regardless of whether the comments are accurate or not, can powerfully influence a company’s prospects for success. Consider, for example, the impact of a bad review of a hotel, a book, a bank or a recently launched laptop.
The challenge is how to harvest quantifiable insight from the mass of unstructured data (commentary) that resides on the web. When dealing with popular brands the volumes of comment are naturally massive and the ability to decipher complex sentences is not straightforward. Of course there is also the not so trivial task of identifying which sites to harvest in the first place.
A variety of search engines have appeared which allow simultaneous searching of multiple sources of social media commentary. Although these are quite basic and provide a limited universe of the available comments, such as, types of sites to include in the search, geography and time periods they provide an early indication of the nature of the commentary – generally good or bad.
But how does one turn vague expressions of sentiment into worthwhile structured data to combine with financial indicators? The answer is with some difficulty.
Some people express sentiment in black and white terms. Bad is bad and good is good. But how bad is “bad”? And, is good just good, extremely good or excellent? How does one reconcile or moderate one person’s assessment of good against another’s. And how does one deal with a comment that “product X is not that bad....” Is an automated algorithm that quantifies sentiment going to just pick up the word bad or is it going to be clever enough to recognise that “not that bad...” may actually be good!
Add to this the infinite possibilities of nuance in different languages and cultures and the scale of the problem of turning social media into useful non-financial indicators becomes readily apparent.
But we should not give up, for we are at the beginning of a very interesting journey. The search tools are primitive, the databases are truly massive and the algorithms are complex but the rewards for capturing and utilising sentiment in business forecasting are potentially very rewarding.
