Data doesn’t have to be perfect – just useful. Depending on the size of your organization, you could save thousands to millions with these techniques.
CASE STUDY – An equipment repair service company is losing money and they suspect the main issue is pricing. They’re tempted to raise all prices but it’s a highly competitive environment and they’re afraid of losing complete contracts. They need better data, but they don’t have a system that tracks time and repair information for each piece of equipment or service provided. The problem festers and profits suffer.
Here are some steps to improve data accuracy and data analysis.
- Determine the accuracy required for the problem at hand. Life and death medical data analysis probably requires more accuracy than the pricing of your products and services. Consider the cost of failure or being wrong.
- Fix the data if possible. Solicit staff input and understand data accuracy concerns and causes. Take action, such as cleaning up the data entry issues, “scrubbing out” duplicates or bad entries, fixing the technical glitch, conducting training to emphasize accuracy and input methods, etc.
- Determine if the data is actionable. Compare the total cost of improving accuracy further – and the cost of your problem – against the potential return of better management information.
- Create an analysis model. In this case, we’re looking at pricing, based on historical activity.
– Make assumptions. Could this equipment repair service use a “best estimate” of 15 minutes to fix machine “A,” 30 minutes for machine “B”and 1 hour driving time round trip?
– Refine your assumptions. How far off could you be? What’s the range of time it takes to repair? Who’s your best and worst technician? Your oldest and newest machines? Are some territories more spread out geographically than others? Does one machine fail twice as much as another? Could you refine your estimates based on this knowledge?
– Analyze a sample set of data. If it takes too long to analyze a year’s worth of good data, could you work with a 1 month or quarter sample? If you don’t have data at all, could you collect some for a week, month, or quarter to get a representative sample?
– Test your assumptions for impact. What’s the impact of the potential “bad estimate” vs. the impact of the potential return? Most people don’t realize that having many variables in a data analysis model often minimizes the impact of some of the variables individually. In this example, let’s assume overhead figures and parts costs are known somewhat. When you do the math, you’ll quickly find that a labor variation of 10 minutes or so may not change the eventual outcome of your pricing decision as much as you think. Then again, it may be more than you think. Either way, your financial model will help distinguish between unjustified fears and safe decisions. This helps create one version of the truth for your team. Your model could also highlight other profitability root causes.
– Mitigate other risks. You can gain more confidence with your proposed decision through competitive intelligence, staff discussion, and customer dialogue. You could pilot the price changes in one territory or on one line of equipment before full implementation.
Respect data accuracy, manage it, mitigate it, and move on with your business. Don’t let data accuracy and data analysis issues stall your improvement efforts.
Edward Livesay
