Editor’s note: Tim Macer, managing director of meaning ltd., writes as an independent software analyst and advisor.

It’s been quite a wait, as even the people at SPSS admit, but the arrival of Dimensions 3.0 in May adds several more vital pieces that were missing from this particular jigsaw puzzle. The grand project is not yet finished, but it now looks like a recognizable picture with a few gaps in it, rather than a handful of puzzle pieces optimistically linked together before you’ve even got the border set out. The two big advances of this version over its predecessors is a fully-fledged and Web-enabled end-user interface - giving the Dimensions software its own front door and its own customer-friendly reception area for the first time - and support for CATI. Until now, Dimensions was Web- and paper-only; for CATI you needed to use one of SPSS’ legacy tools, chiefly Quancept or Surveycraft.

The sheer scale of the Dimensions initiative demands awe, if not respect - it is clearly the most costly and probably the most complex piece of MR software ever developed. It started life as Vision 2000, which SPSS announced at its 1999 user conference - a name that has come to haunt it in the subsequent six years.

But while other software companies were producing complete, menu-based systems where syntax and scripts were banished into a rarely-visited back room, early Dimensions software, by stark contrast, seemed to be getting complex and more technically demanding. This strategy appealed to the larger research companies with the developers in-house to tame this beast, but left the SMEs and the middle ground scratching their heads. Significantly, it left North America ’s backbone of fieldwork houses and in-house users out in the cold, looking for a way in.

In from the cold

Version 3 flings that front door wide open and it is the recent upgrading of DimensionNet which makes the difference, by banishing the complexity to the back room again with an approachable and user-friendly front-end for researchers and non-technical users.

DimensionNet, although largely the subject of this review, is not actually a product at all, if you define a product as something you pay for. You get it free as a unifying wrapper around any of the Dimensions “mr” modules, such as mrInterview, mrPaper, mrTables or mrTranslate.

It gets off to a good start by presenting a very helpful library of all your current and past projects instead of having to hunt through folders on shared drives to locate specific files. Some potentially disastrous consequences awaited naive users who messed with the files in these folders. Now these can be controlled centrally and hidden from view to all except the few who need to mess.

The most obvious innovation in DimensionNet is its new questionnaire authoring tool. The interface design is pleasing and uncluttered. You work on a kind of mode-neutral questionnaire, because in Dimension 3 you could be interviewing simultaneously on the Web, on paper and on CATI (more on that later). In future, other interviewing modes will come too. The developers seem to have applied the 80/20 rule to the kinds of questionnaire constructs you are allowed to create directly from the questionnaire editor. Most everyday things are possible but there are still a lot of irritating restrictions that I hope future releases will resolve. For example, you cannot perform any calculations on your questions, and once you have added sample, you cannot retrospectively apply any filters to it.

Is that all it does?

Overall, the questionnaire-builder tool is beguilingly simple, to the extent that you can see all there is to it in a couple of hours and find yourself thinking, “Is that all it does?” But of course, there is more, hidden from view - all those mind-boggling things the Dimensions developers at SPSS have been working on for six years, in fact.

No tool can hope to do everything, and those that try to end up being difficult to learn and no longer well-aimed at their target audience. It is a tension that the DimensionNet approach manages very satisfactorily. In between tasks, even in between questions, you can introduce snippets of Dimensions script, which is SPSS’ special Visual Basic-like language for market research activities, or even real VB. So everyone is happy: researchers can focus on questions, analyses and research issues in their design without having to become programmers, and the technical people are still in demand to write the clever stuff, but without the boring bits. That is the plan, anyway.

Once a script has been created it can be stored in a library for anyone to call up. It is just a pity that scripts always rely on a small element of syntax to kick them off, and could not be invoked by non-techy users through a button or drop-down menu.

Unusually among its peers, Dimensions incorporates complete version control so that you can both change questionnaires at any point and also roll back to a previous version if you need to. It is an impressive and much needed feature. Furthermore, multiple waves of the same study can be consolidated automatically and there are also simple, error-free tools to help link unrelated studies with common questions or components into a composite view for reporting. While Dimensions still takes a survey-centric view of research, these capabilities are helping to flex the definition of what a survey is, deal with trackers and create meta-surveys from existing and unrelated datasets.

And CATI made three

MrInterview is a from-the-ground-up multimodal data collection product, which until Version 3 was fine if you only ever wanted to combine Web and paper surveys. Mercifully, Version 3 also introduces the support for CATI that has been stretching the patience of Quancept and Surveycraft customers to breaking point. Or rather, it almost does.

At the authoring stage, CATI is simply another mode that you can apply to your survey, and at the execution stage, mrInterview now contains a broad range of grown-up CATI features such as centralized sample allocation and distribution, user-definable call-back rules, integrated quota control. Everything is very flexible, so all the rules can be changed, as can all the assumptions about the roles and permissions of supervisors and interviewers. You could create your own subgroups of both and give them enhanced capabilities, or clip their wings for safety’s sake. And most helpfully, all the menus of options and rule sets are populated with carefully thought out default sets to deal with the “80 percent” of what people do in CATI anyway.

While increasingly the other MR software providers have tended to go for standard database reporting tools such as Crystal Reports to provide system and survey management reports - which means investing in another set of skills and tools - Dimensions cleverly uses its own reporting tool, mrTables, to deliver real-time reporting to supervisors through their DimensionNet Web interface. Again, a slew of “standard” reports are provided, but these can be changed or supplemented with your own reports, which any tab specialist will be able to write. One current lack is record-level reporting, to examine individual responses, which needs to be addressed as a priority.

Quancept users can migrate gradually to CATI in Dimensions, as their existing Quancept scripts can be used with mrInterview, but Surveycraft users will be disappointed that this luxury does not extend to them.

For serious CATI users, the new mrInterview probably contains a couple of problem areas that SPSS will not be addressing until a subsequent release, notably integrated telephony, for dialer and also voice recording, and the fact that interviewers need to use the mouse to select responses, rather than the time-honored numbered code punched in on keyboard or keypad. However, that will be fixed later this year, as SPSS makes the interviewer interface Section 508-compliant, which will mean being able to perform all selections from the keyboard or the mouse.

Far-reaching

The system is totally Web-enabled as it works within the Microsoft .Net framework and is therefore location-independent, enabling work to be shared between offices, enterprises and outworkers. It is a true multi-user system and impressively manages all the contention that could arise if two or more people try to work on the same survey at once. They can, but only one user can make changes, even through others can be viewing at the same time.

The same administration tool that defines roles for interviewers and supervisors lets you define users and permissions for researchers and clients too. You can use this to provide logins for clients to view results or to specialists to perform particular tasks such as entering language translations. It is a pity that you cannot customize the appearance of the client access pages: They remain branded as a free ad for SPSS.

Also on the downside, testing routing and previewing a questionnaire is clumsy, which slows down error detection and the final polishing stage considerably. It is also unfortunate that the documentation for this and all SPSS software now appears principally in the form of online help. I miss the good read during which you have a chance to deepen your knowledge and expertise. And despite everything, this tool still only supports two of the four common interviewing modes: paper and Web.

Won the show

Todd Myers, vice president and director of technical services at Opinion Research Corporation (ORC) in Princeton , N.J. , has recently taken his firm through detailed software review, with a view to replace two disparate data collection tools for Web and for CATI with one multimodal solution. “SPSS won the show,” he says. “What got us was the Data Model and all the efficient ways it offers to manage all the research data we collect. We have struggled for a long time to store categorical data in a database and access it efficiently. The Data Model provides us with a way to transform and access the data without starting from scratch with our own programs.”

ORC decided to wait until 3.0 was released, for both the CATI and overall interface and ease of use improvements in the new version. “The features in general are extremely strong,” Myers says, singling out Dimensions scripting, which he sees providing the ability to automate much of the research process and “pump things through from data collection, through cleaning and out to reports or, these days, the Web.”

SPSS has “have done a nice job of building a VB-like scripting language that is aware of market research,” Myers says. It was, however, a concept that was more readily understood by those in his team with a database or programming background, but which, initially, some of his most experienced script writers, with 10 or 20 years behind them, found a struggle and even a worry to grasp. After a specific session with SPSS aimed at this group, they were able to understand it and see how they could benefit from this new approach.

“Their learning curve will be a bit slower, but I have not seen anything that says they will not get there. They see the ability to automate scripts as a real advantage,” Myers says. However, it does mean that some rudimentary database and programming skills will be featured in ORC’s person spec for future hires to the scriptwriting team.

Dream is reality

Another break from the past is the cost. SPSS has revamped its price model and now offers affordable entry-level pricing to attract back the SME and in-house users it seemed to have deserted for a while. I have already seen this make Dimensions much cheaper than its high-end competitors, who have been enjoying a few good years while the Dimensions vision lay over the horizon. The dream is reality. It’s not a vision anymore.