Research Methods
Research Methods for Business Students, Sixth Edition, by Mark Saunders, Philip Lewis, Adrian Thornhill, Pearson 2012About a month ago we were shown the results from a very pricey research study that was conducted by a firm working for one of the world’s tech giants. The study was supposed to look at IT usage in a particular industry segment. The observations and conclusions were intended to guide future strategic direction of IT market development, as well as demonstrate from real-life examples, where firms have explored new and better IT solutions. The survey cost well into six figures, produced a white paper and various written results, but it failed.
How could a timely, much-needed survey on such a critical topic fail to deliver? Despite pages and pages of interesting (should we call this anecdotal) corporate responses, the sample size was too small. As The Mill Girl learned at Rath and Strong, where Dorian Shainin was the uber-guru of statistical work, a sample size under 32 is questionable. Given best case opportunities for sampling a large population, that number should not be impossible to reach. But with a new, somewhat controversial subject whose results should be based on real corporate experience, things get harder.
Some companies don’t have the data, or they don’t have time to dig it, and some companies have great data, but they don’t really want to share it. And in the era of Internet-based research, where so many hits don’t guarantee valid responses, it’s difficult and probably too expensive, to cultivate the field. Where does that leave the business researcher? Making-do, that’s what.
But numbers have great power. They direct us to ask the right questions, as well as providing actionable answers. A small selection of fully-disclosed, albeit disguised case studies, has long legs, value beyond the question of the moment.
What are the kinds of questions that one might use to develop meaningful surveys? How about an analysis of customer demand that depletes your warehouse and launches hugely expensive manufacturing speed-ups or expedites? Or we might look at an entire population of automotive suppliers to better understand how well they can ramp up their production, and in which areas one can expect hesitation or bottlenecks. IN the Information Technology area we might look at Cloud adoptions and legacy system disposition – parallel implementation and sun setting, cold turkey phase-ins, or a third very common category, “I don’t know.” It’s amazing what we can learn from a well-designed simple survey.
My first surveys were done with histograms and eye-balling shopping carts filled with “hernia reports.” Once we had developed a “feel” for the data, it made sense to sharpen the corners a bit and start to structure the results. I learned from Clint Jones, an outrageous and brilliant Rath and Strong consultant, to always structure the survey to get to the results you want to uncover. Big blind surveys just weren’t what we needed then. As my tools improved handling more and more data did not always give me an advantage in running the research - something like drowning in a sea of numbers, I discovered that once the web kicked in, it was too easy to be suffocated by the very numbers we wanted to play with. Hence the emphasis on survey design, picking the right population, deciding whether to work with a completely random sample, or what we called at Rand and Strong, “a stratified random sample.”
Chapter 7, “Selecting Samples” in Research Methods for Business Students, hits this point dead on. Working with original client data has advantages and disadvantages – client data has the most meaning and relevance, but it may also be overwhelming. On-line surveys require a high response rate to produce good, actionable results or questions, but users need to understand the questions below the surface of web vehicles that make it all look so easy, including the levels of non-response:
· Complete refusal: none of the questions answered;
· Break-off: less than 50 per cent of all questions answered other than by a refusal or not answer
· Partial response: 50 per cent to 80 per cent of all questions answered other than by a refusal or no answer;
· Complete response; over 80 per cent of all questions answered other than by a refusal or not answer.
This volume is used by MBA and other students developing dissertations and research projects. It is so well-structured and offers a companion website that I found is a wonderful review vehicle. The website www.pearsoned.co.uk/saunders offers self-test options, as well as tutorials and datasets for Excel, NVivo and SPSS. There are additional case studies to sharpen your problem formulation skills, an online glossary, and a most useful new tool, the Smarter Online Searching Guide, designed to help researchers made the most of the internet in their research.
***