Time is of the essence when tracking the effect of events in both continuous and ad hoc research. However, the sheer effort and complexity of delivering sample to fieldwork agencies within hours makes this an impossibly expensive task for most companies.
The information needed to is often tantalisingly close, but when you take into account the many different databases and database tables that you need access to, then consider how to ensure samples are both random and representative, you soon have a major IT project on your hands. And that is before you consider niceties such as ensuring your most valuable customers arent over-researched, requests not to be contacted by phone and opt-outs are honoured and that the sample derived falls perfectly into the sub-groups needed.
MaRSC, from UK-based Centurion Marketing Systems, is as close to an off-the-shelf software solution you can get for this loosely defined activity. Its acronym stands for market research sample control. It does this by acting as a bridge between corporate databases and data warehouses and market research data collection systems, and by being an intermediary between the research buyer and research agency. Sample can be delivered to agencies through extranet links, automated FTP transfer or emai and can be issued as fixed format ASCII files, or in various metadata formats, including XML and triple-s.
An unusual but highly valuable by-line of this product is that it allows research buyers to compare the performance of fieldwork agencies. Once interviewing is over, the back-feed, when the outcome of the interview is applied to the database, and non-contact can be identified and released for other projects, can be reviewed. This review can show fieldwork hit rates and MaRSC will even take this history into account when generating new samples, adjusting the quantity of sample for the measured efficiency of the agency. Interestingly, corporate users report great reluctance on the part of research agencies to provide this information.
As for the product itself, at the front end there is an easy-to-use set of Windows tabbed menus where you define the characteristics of your sample, set any filters or suppression criteria, and define the quota targets in absolute or percentage terms. You then tell the program to generate a trial report-only run. Any shortfalls colour-coded in the report for easy recognition.
Drop-down lists, built from the database tables, make it easy to define your sample and set up your quota tables, which can interlock down to seven different levels, Sophisticated and secure sample resting rules will ensure your respondents are not over sampled - no matter what level of desperation the requestor has reached.
MaRSC can either sit astride its own, dedicated relational database of sample, or it can be configured to avoid needless duplication of data and synchronisation problems by utilising an existing data warehouse of customer records. To achieve this, several MaRSC-specific tables need to be added to the database to record sampling and interviewing history, which can be a step too far for some database managers.
|The program will happily draw sample as a random selection from the database it controls, but another great feature is its ability to use an external file to drive the selection process. It means you can supplement your own customer database with other bought-in lists or samples for ad hoc research projects, and you can generate timely sample for event-driven research.
One point to watch is that MarSC it is a customised rather than off-the-shelf solution, and for event-driven research this involves several other standalone upstream and downstream modules. Its rather a pity that these all stand outside the main GUI interface and would be better to have more of a workflow approach.
Robert Jacob, Head of Research Systems at Barclays Marketing sees key benefits in MarSC, where it is used across the entire Barclays UK consumer operations, in being able to target sample better and maintain a company-wide research contact history. Very few companies take the issue of sample management seriously, he says. Yet it is incumbent on us all to maintain a proper contact strategy or we risk damaging the reputation of research through too many contacts that are just not relevant or interesting for the respondent.
Jacob has, on occasion, drawn sample with as many as seven different interlocked target levels. Of course, it depends on the information you have, but what we are finding is that if you use everything intelligently, you can often use known attributes as proxies for unknown ones. And it works - the success rate we get in obtaining a response is far higher when you use this information to create highly targeted samples.