`Home     Our Services     Your Resources     News & blog     About us     Contact

MarketSight reviewed

Tim Macer reviews a package that allows you to simplify and refine the process of getting the most of your data

One of the poorest-served areas of research software is in the arena of analysing and interpreting results, especially for end-users. There is certainly plenty of choice, but few make the effort to build knowledge into the tools they offer, such that the path to safe analysis becomes the obvious one. MarketSight is a rare exception. It recognises the steps that people typically follow when approaching their data, and simplifies and refines these into a tool that encourages you to venture further into your data, to dig up connections and findings you would not otherwise realise were there.

The tool is provided as a 100% web-based product. It requires a PC running Windows and Internet Explorer 6, so does not offer any cross-platform capabilities. It is offered as a hosted service or as an enterprise product, to install on a company’s own web servers. Though it is very easy to pick up, especially if you have some familiarity with other data analysis tools, it is very different in a number of respects, particularly in the area of hypothesis testing through the use of significance tests, where it operates as closely to as an expert system as anything I have seen.

The software recognises that there are usually two phases in working with your data, once you get your hands on it - to start with the expected then move on to the unexpected. To handle the first stage, MarketSight offers a ‘banner’ tool, which is ideal for checking results and getting an overview. This effectively lets you create a complete set of cross-tabs using a standard break (or banner, as Americans refer to them). It is simply a matter of ticking a few boxes to select some or all of your survey questions and pick some break questions.

To save waiting, the process runs in the background, though it usually takes just a few seconds to complete. Very sensibly, the output is directed to an Acrobat PDF, which is bookmarked, so it is easy to zip up and down the file, and is printer-friendly too. Presentation options are limited in this quick snapshot view, but significance tests are applied to the results and significant values are highlighted in colour. You can produce enough to keep you busy for many hours in just a few seconds.

To dig deeper, you use the custom analysis tool which lets you hand craft individual cross-tabs and apply a wide range of presentation options, but again through a simple checkbox-style selection method. At this point, the ‘expert system’ capabilities start to appear, in that the software decides for itself what is an appropriate significance test to apply according to the context. Each test has its own colour-code, and when a colour-coded cell is visible. If you mouse over it, a tool tip will appear presenting a short explanation of the test applied, its parameters and the level of significance. Furthermore, whenever you are in the analysis tool, a panel of context sensitive hints and tips are visible to the left of the main window, offering a layman’s explanation of any stats applied and other guidance on analysis and interpretation. You can alter the significance level or apply error correction, and if you want, you can filter the output only to show results that are significant (not an approach I am usually comfortable with).

As you would expect, the tool lets you create and apply filters, apply (but not create) respondent weighting, create new variables and combine existing ones, and even recode a whole series of variables automatically once you have defined a model to follow: potentially a real time-saver. However, one major deficiency, which could have you stop reading this review at this point is its lack of support for multi-coded variables. This will be rectified in the future, but it does pose some interesting challenges on the sig testing side.

Data very are easily imported - provided you start off with SPSS or SAS data, when all the variable names and labels will appear. There is, alas, no triple-s import (but then it cannot do anything with multicoded questions yet). Excel is the only other option, and that involves a lot more set-up to get the texts and codes right. However, a very intelligent feature is the wizard to help you combine successive waves of continuous data, which it will largely match automatically and ask you only to resolve any differences that emerge. All of the likely differences can be accommodated easily with a few mouse clicks, making the tool ideal for work on trackers.

The help and documentation provided online is of a high standard of clarity, explaining advanced concepts in ways that anyone can understand. It is model in user empowerment. Support is provided from Cambridge, Massachusetts, though Monitor does have an office in London, and pricing is in US Dollars.

The User View: Boston Scientific

Kim Goldman is Senior Marketing Science Manager at Boston Scientific, one of the world’s largest providers of cardio-vascular medical devices and a vast range of other non-invasive medical devices.Her department provides strategic guidance and support in marketing to the company’s many product divisions, each of which is responsible for its own marketing. As a part of that,Kim and hercolleagues carry out a large amount of marketing research and have been using MarketSight to analyse the data for a little over a year now.

“In the past, we largely relied on Microsoft Excel and to some extent on SPSS,” Kim explains. “Analysing surveys in excel is a very time consuming process. About the same time, we grew our department from 5 people to 30, and started performing more analyses with a greater degree of complexity - leading to a need for something much more powerful.”

“Evaluating segmentation and significant differences between customer groups is key to many of our analyses. Doing that in Excel is pretty close to impossible. MarketSight saves us time and allows our team to test multiple options that often reveal unexpected findings. Because the statistical significance is built in, you can decide immediately whether or not these segments are truly behaving differently.”

“If I were to point out one feature in this software, it is the way that MarketSight manages statistic significance. That’s a very complicated subject for most non-statisticians, myself included. The way it is set up - with highlights and commentaries on what the statistics mean, how they define the different shading and what it is telling you - has proven to be very useful to people in our department.”

MarketSight also frees the team from having to use specialist IT or DP help to get projects set up. “If I am using an SPSS file, which is the only format I use now, it is the easiest method to get import all of the data into MarketSight,” Kim reports. “We have never had a problem with a data upload. We then create the groups, so all the questions are combined on a topic basis.”

Kim considered the tool was reasonably easy to use in the previous version without formal training, but has observed her colleagues have become even more productive with version 5.0, due to several time-saving additions. These include systematic regrouping of answer lists, charting and the new ‘banners’ survey snapshot reports.

Published in Research, the magazine of the Market Research Society, June 2006, Issue 481

© Copyright Meaning Ltd 2006. All rights reserved. Reproduced with permission.

Web-based survey analytic tool for researchers and end-users, with an emphasis on interactive hypothesis testing.

Ease of use
Cross-platform compatibility
Value for money

Hosted solution starts at $1200 per user per year, with multi-user discounts available..