What experts say about Information Quality: The Potential of Data and Analytics to Generate Knowledge:

A glance at the statistics shelves of any technical library will reveal that most books focus narrowly on the details of data analytic methods. The same is true of almost all statistics teaching. This volume will help to rectify that oversight. It will provide readers with insight into and understanding of other key parts of empirical analysis, parts which are vital if studies are to yield valid, accurate, and useful conclusions.

There is an important distinction between data and information. Data become information only when they serve to inform, but what is the potential of data to inform? With the work Kenett and Shmueli have done, we now have a general framework to answer that question. This framework is relevant to the whole analysis process, showing the potential to achieve higher‐quality information at each step.

The authors have a rare quality: being able to present deep thoughts and sound approaches in a way practitioners can feel comfortable and understand when reading their work and, at the same time, researchers are compelled to think about how they do their work.

No amount of technique can make irrelevant data fit for purpose, eliminate unknown biases, or compensate for data paucity. Useful, reliable inferences require balancing real‐world and theoretical considerations and recognizing that goals, data, analysis, and costs are necessarily connected. Too often, books on statistics and data analysis put formulae in the limelight at the expense of more important questions about the relevance and limitations of data and the purpose of the analysis. This book elevates these crucial issues to their proper place and provides a systematic structure (and examples) to help practitioners see the larger context of statistical questions and, thus, to do more valuable work.

…the “Q” issue is front and centre for anyone (or any agency) hoping to benefit from the data tsunami that is said to be driving things now … And so the book will be very timely.

Kenett and Shmueli shed light on the biggest contributor to erroneous conclusions in research ‐ poor information quality coming out of a study. This issue ‐ made worse by the advent of Big Data ‐ has received too little attention in the literature and the classroom. Information quality issues can completely undermine the utility and credibility of a study, yet researchers typically deal with it in an ad‐hoc, offhand fashion, often when it is too late. Information Quality offers a sensible framework for ensuring that the data going into a study can effectively answer the questions being asked.

Policy makers rely on high quality and relevant data to make decisions and it is important that, as more and different types of data become available, we are mindful of all aspects of the quality of the information provided. This includes not only statistical quality, but other dimensions as outlined in this book including, very importantly, whether the data and analyses answer the relevant questions

This impressive book fills a gap in the teaching of statistical methodology. It deals with a neglected topic in statistical textbooks: the quality of the information provided by the producers of statistical projects and used by the customers of statistical data from surveys, administrative data etc. The emphasis in the book on: defining, discussing, analyzing the goal of the project at a preliminary stage and not less important at the analysis stage and use of the results obtained is of a major importance.

Ron Kenett and Galit Shmueli belong to a class of practitioners who go beyond methodological prowess into questioning what purpose should be served by a data based analysis, and what could be done to gauge the fitness of the analysis to meet its purpose. This kind of insight is all the more urgent given the present climate of controversy surrounding science’s own quality control mechanism. In fact science used in support to economic or policy decision – be it natural or social science ‐ has an evident sore point precisely in the sort of statistical and mathematical modelling where the approach they advocate – Information Quality or InfoQ – is more needed. A full chapter is specifically devoted to the contribution InfoQ can make to clarify aspect of reproducibility, repeatability, and replicability of scientific research and publications. InfoQ is an empirical and flexible construct with practically infinite application in data analysis. In a context of policy, one can deploy InfoQ to compare different evidential bases pro or against a policy, or different options in an impact assessment case. InfoQ is a holistic construct encompassing the data, the method and the goal of the analysis. It goes beyond the dimensions of data quality met in official statistics and resemble more holistic concepts of performance such as analysis pedigrees (NUSAP) and sensitivity auditing. Thus InfoQ includes consideration of analysis’ Generalizability and Action Operationalization. The latter include both action operationalization (to what extent concrete actions can be derived from the information provided by a study) and construct operationalization (to what extent a construct under analysis is effectively captured by the selected variables for a given goal). A desirable feature of InfoQ is that it demands multidisciplinary skills, which may force statisticians to move out of their comfort zone into the real world. The book illustrates the eight dimensions of InfoQ with a wealth of examples. A recommended read for applied statisticians and econometricians who care about the implications of their work.

Kenett and Shmueli have made a significant contribution to the profession by drawing attention to what is frequently the most important but overlooked aspect of analytics; information quality. For example, statistics textbooks too often assume that data consist of random samples and are measured without error, and data science competitions implicitly assume that massive data sets contain high‐quality data and are exactly the data needed for the problem at hand. In reality, of course, random samples are the exception rather than the rule, and many data sets, even very large ones, are not worth the effort required to analyze them. Analytics is akin to mining, not to alchemy; the methods can only extract what is there to begin with. Kenett and Shmueli made clear the point that obtaining good data typically requires significant effort. Fortunately, they present metrics to help analysts understand the limitations of the information in hand, and how to improve it going forward. Kudos to the authors for this important contribution.