Cost reduction programmes rely on fast, effective decision making, which in turn relies on high quality data. The education and practise of managers, whether PPM or finance managers is focused on a broad range of frameworks, processes, tools and techniques that explain the world and describe seemingly credible ways of optimising any project or business. Tremendous effort goes into detailed analysis of say a planning network chart, an annual budget, a project business case or cost reduction plan, with a focus on optimising any particular situation. Yet these standard business, finance, and programme management analyses assume full availability of data, the existence of an optimum solution, and full rationality. This neatly sidesteps an inconvenient reality:

The data is often poor and misinterpreted
The environment is often so complex and unpredictable that an optimum solution may not be obvious or even exist
Even where an optimum solution does exist, there are many psychological traps that prevent the average decision maker finding them

The next newsletters will cover each of these issues in turn.

DATA: The data available will often be inaccurate, out of date and misleading

It is 4pm. You need to provide an initial cost reduction analysis report for the programme Board tomorrow at noon. The untouched email with the data download from SAP is in your inbox (as it has all week). Clearly this effort is doomed to fail, but even if you had been using this data all week, there are a number of key considerations when gathering and analyising data that tend to get overlooked. Far too often, limited data is taken at face value, analysed and the result communicated, with far too little effort in defining the question, cleansing the data, understanding the context, or putting the final 10% of effort in to communicate the results in a powerful way. In my experience it is astounding that the moment numbers are on a spreadsheet, particularly if sourced as a download from an accounting system, they are seen to be a pure, unquestionable and complete representation of the truth. Key considerations include:

Definition	Simple issues can often prevent direct aggregation and comparison. Eg: Is the accounting system on a 4,4,5 week, calendar month, or 13 x 4 week period basis? Do sales include VAT, intercompany sales, retrospective discounts? Categorisation: is the Sales Director or telesales telephony cost under Sales or Central? Are corporate or intercompany services charged at all, at cost, at standard cost, or at market rate. When aggregated, are costs fully costed or marginally costed? The question “what does X cost?” can be one of the most difficult questions to answer, with maybe six different answers depending on how you define cost. I have seen customers exited because they are 'unprofitable', despite making a healthy contribution to bottom line profit; departments 'benchmarked' and consequently downsized; or projects incorrectly cancelled because of 'insufficient' ROI, all because apples were compared with pears. Always ask what was the source and definition of the information. Be nervous when data is compared across different divisions/companies, or sourced from different accounting systems.
Completeness / Breadth	Does the data collected provide sufficient breadth of information to cover the question / problem in order to support the decision making required? Don't just get accept what you are given.
Accuracy	Does the data have the necessary level of accuracy for the decision at hand? An upfront sizing up of a problem, or filtering of potential areas of focus, will need a much lower level of accuracy, than a detailed analysis. The data simply needs to be sufficiently accurate and complete for the specific. Finally, be wary of one-off adjustments in the financial data. There can often be a number of one off charges or accrual releases that confuse the underlying picture.
Precision	Distinct from accuracy (imagine a thermometer that can measure +/- 0.1oC but mis-calibrated so it overstates temperature by 10oC. – precise but not accurate.) Analysis is often grossly over stated in its precision – if the assumptions in a calculation are accurate to +/- 10%, how can the result be represented to the penny? Of course maintain precision of input data through the calculation to allow identification and matching of data sources; risk assessment, but adjust when outputting the results.
Timeliness & perishability	With a general increase in the rate of change, the importance of data that is up to date is growing. A judgement may be necessary as to whether information is passed its sell by date.
Contextual Data	To understand data and separate out the important from the unimportant, it is important to understand the environmental, structural, service changes due to reorganisations, new customers, new products etc. If there has been major underlying change, or one off events, make sure the data is normalised and/or graphs are annotated, and commentaries provided.
Depth / Granularity	Always seek data one level of granularity beneath what you think you need. This allows you to understand the data better and identify any issues. Once aggregated you lose the ability to understand the source of the uncertainty.

Always plot data, even if not for presentation, to check for obvious errors in data.

Beware of characteristics changes driven by accounting policy.

Impact of 4,4,5 accounting periods. To correctly set trends, need to restate 5 week periods as if 4 week periods.

Cost reduction 3 characteristics driven by accounting policy Cost reduction 4 characteristics driven by accounting policy

Next time you see a stated relationship, if they don't show the underlying data, be wary. You need to know the correlation and pattern to make a judgement on the accuracy and validity of the stated relationship.

False precision

Management Consultant: How many sheep do you have?
Farmer:1003
Management Consultant: That is very precise. How do you know?
Farmer: About a thousand in that field and three have escaped into that one.

Ten steps to effective data quality:

Be clear on the question you are trying to answer. What is the hypothesis you are seeking to prove? Know what you are going to do with the answer.
Gather contextual information, quantitative and qualitative, and hone requirements.
Gather core data, but look to get greater set of data; broader and deeper than you think is required.
Make sure you are clear on the precision, accuracy and perishability of the data.
Review implicit and explicit assumptions, definitions, policies, etc, underpinning data.
Plot raw data to check for anolomolies.
Undertake appropriate data analysis.
Sense check results with intuition, previous analysis and with crude independent analysis.
Undertake sensitivity and breakeven analysis. Ask yourself what would need to change in the inputs to change the answer, how does the input sensitivity compare to input accuracy? How confident can we be in the results?
Take care in the presentation of the results. Be clear, graphical and smart. All too often significant effort in data gathering and analysis is let done by a cursory effort in the presentation of the results.

The latter point, could be an article in its own right. At this point I shall simply point out some of the typical pitfalls in effective communication.
Insufficient data and analysis, or conversely data overload with pages of tables of data.

Too little white space – dense pages of text. Difficult to read and assimilate.
Reports cover multiple pages, with no summary.
Supplying inadequate context for the data, underlying assumptions or confidence limits in results.
Displaying excessive precision eg telesales costs, £718,276.45: £718k.
Using poorly designed display media, distorting figures, and highlighting important data ineffectively.
Cluttering the display with useless decoration, or overusing colour