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


New power and flexible friends for cross tabbers

SPSS 7.0 and ESPRI, March 1996

Tim Macer checks out the latest in stats packages for PCs - and finds reports of an old warhorse's demise premature

The notion that that cross tabs, as we know them, are about to go into terminal decline has been in circulation for years.But the future is never so radically different - look no futher than two of the newest releases in the stats packages scene.

In January, SPSS launched version 7 of their well known and highly respected statistical analysis system. A few years back, SPSS was starting to look a bit long in the tooth. This all changed with version 6 and a wonderful new interface for Windows, making it so much simpler to select and view statistics and charts on screen. Version 7 effectively completes the task by improving the presentation and output side of charts and, rather interestingly, tables. If you thought SPSS was great for stats but let down by rather scrappy tables, then think again.

SPSS 7 introduces "pivot tables", a feature offered by several other interactive tab programs on the market. You choose what goes into your table - into the rows, the columns and the third dimension, which they call a "layer". You view the table on screen rather like you would a spreadsheet, but icons to the left of the display represent what is in your "layer", icons to the right, what makes up the rows and along the bottom, the columns. You can drag them from one place to another and immediately "pivot" your table a different way so the columns become rows, or the rows become layers and so on. As a data exploration tool it is very compulsive. A couple of things make their implementation of this especially clever. First, little arrows on either side of the layer icon allow you to flick through the layers. If you choose to put a variable such as ITV Region in the layer, then you will be filtering the table successively by each different region. Better still, your icon could be a presentation option. Each layer could be, progressively, an unweighted total, a weighted total, a percentage, summary statistics and significance values. For instance, you could browse through looking only for significance, then quickly flick back to check that your bases are fine.

Support for presentation options is the best I have seen to date in any tab package. It is easy to forget that SPSS, first and foremost, is not a tab package at all. Using their new "table looks" feature, you can define a style rather like you might in WordPerfect or Word, to give consistency or apply a house style to report headings and paragraphs. In SPSS 7, a "table look" does the same for each component of a table or chart. It comes with a mixed bag of named pre-defined styles, from the restrained "Academic" to the migraine-inducing "Hot Dog". You can add your own too.

SPSS still retains the concept of a script which you execute, and which originally you had to write yourself using arcane mnemonics and codes. Now you need know nothing about this, as you can complete everything using menus, pop-up lists and check boxes on screen. This does mean it is easy to set up standard reports on one wave of data, then run the same reports on each subsequent wave. When you save your work, the script is saved ready for you to run again, or to modify or add new requirements. All you have to do is remember to hit the run button, otherwise you will still be looking at last month's data. It is as simple as that.

For most people, the difficult part of a stats package is understanding the stats themselves, or selecting appropriate ones to use in different contexts. The worry for most people is that they will present something using an "eigen value" or a "z-score" and someone in the audience will crushingly say "but that is invalid in this context becauseƒ" While SPSS makes no attempt to provide you with a decision support system, it does now incorporate very comprehensive on-line help and tips about each of the hundreds of statistical tests it incorporates, right down to nitty-gritty detail about which assumptions should be met or when a test is likely to be weak and how to counteract this. This should make it a lot more accessible to the non-statistical end-user they are now trying to woo. John Penn, SPSS's UK Marketing Manager told me "we have tried to make it as easy as possible to get tips on how to, or how not to use the tests so you don't draw the wrong results or conclusion".

SPSS are finding that conversion among their existing users in the USA is going faster than in Britain. Version 7 will only run under Windows 95 or NT 3.5.1. and uptake looks like being slower this side of the Atlantic. However, the writing is on the wall for the old Windows 3 diehards. SPSS could not have brought out this new version without Win95's vastly superior memory and resource management. As developers, they are not alone in this discovery. If, as end users, we want the goods, we will have to bite the bullet, and switch to Win95 or NT.

I spoke to Simon Hearn, Chief Statistician at MORI and an experienced user of SPSS, for his reactions to version 7. "The table pivoting will be of great use, and for people who are not statisticians it is obviously better to have an easier way of doing things. The basic engine underneath is still the same, so there are no real new techniques for the statistician. This release is not so much for people like me but for the exec. who may be numerate and wants to do some cross tabs, then run on into some hypothesis testing and maybe some correlation. They want to be able to click on a few things and quickly get the results they want". Simon was keen to get his new copy, once his machine had been upgraded to Windows NT.

ESPRI, a statistical analysis and data reduction tool from Information Tools Limited is a relative newcomer to the UK even though it is widely established "down under". As Gill Davies, the UK Marketing Manager put ESPRI through its paces, the first thing I noticed was its very comprehensive and accessible support for cross tabs as a means of modelling and viewing data. After that, the similarities between SPSS and ESPRI ended. Gill explained that the product had come about because a group of researchers in New Zealand were looking for ways to eliminate printing endless cross tabs of everything by almost everything else. "They wanted to use a tool to sift through the data and have the interesting findings pop out in front of them." What they produced was ESPRI (an acronym for Easy System for Performing Research Investigation).

Strictly speaking, ESPRI is not a "stats package" at all. It is a tool which uses statistics to allow on-screen data reduction and modelling. It excels in two particular areas of research - segmentation and tracking or continuous research. As a Windows product, ESPRI looks and feels like many other Windows packages, and has all the graphical presentation options you would expect, allowing you to produce presentation quality results in chart or tabular form from your own printer, or cut and paste between other Windows applications. ESPRI's power is in the way it lets you digest your data.

Unlike SPSS, ESPRI is sold as a complete service. Information Tools set up your survey as an ESPRI database, provide you with instruction in how to use the software and include the software tool itself, on a per-survey basis. This works well for organizations or departments who do not have the technical skills to define all the fields and the data, or for agencies who wish to provide something specific to one particular client. One or two high-volume users who do have the skills in house have found it more cost effective to purchase the software outright to allow them to set up their own databases from scratch.

Raymond Marks, MD of London-based agency VA Research, recently supplied results to one of their clients in ESPRI format for the first time. "This is the way things are going", he told me. "Our client found it to be very helpful. It coped with 15,000 cases with no problem. In fact we found it useful to double check our own analyses. Like all these things, you have to put the effort in to learn it - and you have to know which questions to ask it too. But it offers some nice segmentation capabilities, which is a great help. I don't know that it is going to replace the printed table, but for individual analyses it is very useful."

ESPRI offers four tightly integrated modules, the first of which is a reporting module and is where most users start, as it offers easy access to tables and graphs. After that, there is a description tool which allows you to perform data discrimination on your data - identifying the groups you didn't know were there. It is data discrimination that has given us such memorable terms as Yuppies and the Empty Nester. The skill comes in identifiying what it is that makes your groups cohesive.

The mapping module offers straightforward pair-wise comparisons and principal component analysis to allow you to create perception maps on screen in a very easy way. Finally their segmentation tool provides interactive, graphical cluster analysis. They also provide support for mapping using correspondence analysis.

Tania Biggs, Research Buyer at Alliance and Leicester Building Society is an end user of ESPRI. It has transformed the way in which she works. Tania is concerned with tracking awareness and trend data. With about 3 days per month to work on this area, most of her time used to be spent inputting values into charts and tables, and manually extracting data. "Before, I never really got round to doing much reporting or interpretation", she told me. ESPRI allows her to use her time actually making use of the data that have been collected. "I highlight all the questions I am interested in. I then tend to flick through all the graphs [on screen] and can immediately see any changes. It's very fast. I track awareness and propensity to use data and plot this against advertising and promotional data. You can immediately see where there are any blips and you can also spot how the market is moving as a whole".

Having full access to all the data provides numerous other advantages. "Someone would invariably come back and ask us for information on something we had not been tracking - which we weren't able to do. Also, we used to have problems calculating averages of averages". Both are resolved using ESPRI. Tania also liked the ability to form bigger subgroups, if bases were getting low, by combining them within ESPRI.

All of this leaves me wondering what role the traditional cross-tab package will find for itself in the future. The threat does not seem to be that cross-tabs are in decline, more that they are becoming an attractive feature of everyone else's software. Make that two hot dogs for me, please.

Published in Research, the magazine of the Market Research Society, March 1996, Issue number 358.

© Copyright Market Research Society/Tim Macer 1997. All rights reserved. Reproduced with permission.


Margin quote

Copyright © 2006-2009 meaning ltd. All rights reserved.