Datafication is a way of building measurement into your work. Starting that project, you need to do data audit, list your business functions and decision you need to make and connect them with data you need to do them. Then categorize that data. Prioritize the data you need and build ways to get data you don’t have. And at the end find place to put the data.
Almost everything we do today counts either as activity or results. Counting is the first step to saving money. Data don’t make decisions for you, but are good starting point. Productivity is one area where measurement can be done. Second are social media. They are all digital. For the first time in history, our attitudes are datafied as soon as we express them. But be careful, retweets and likes don’t measure how popular you are, whether people will buy something that you Tweeted about, whether people like you in real life. Social data are not data about everything, but it’s the biggest thing free thing in datafication ever and you would be foolish to ignore them.
In 2000 about quarter of all information was digital data. Today 99% of information is digital. Big data is by definition the stuff that is too large and complex for the sort of databases that companies are usually run. Today’s application that employ wisdom of crowds’ type of intelligence has big-data-like properties. For small business using big data is a good investment of they use the for improving their own data. But the problem is that majority of companies even if they do some researches, sometimes they don’t use them to change their business decisions. Big data value is in analysis. But it is good if you start with small projects.
You need to keep your data alive. Collect them in databases and spreadsheets. Not in PDF. Databases are storage for large amounts of data. And data need to be updated regularly. Good table is drawn in a way that important data are kept on top, that we shouldn’t use jargon in descriptions, we should use as few lines as possible, but no less. Use graphs. Graphs can obscure information as well as describe it. Be careful about using 3D and pie charts. They can give wrong impressions or they don’t offer enough information. If you have small amount of data, leave it in a table. Graphs often tell a story about a relationship in data. Identify this story will help you make better decisions. These stories are usually in a format of trend. If you only put numbers into table of graph, this is statistic and they don’t give you full value of information.
One of most used techniques in data research is looking for average. There are three most used ways: mean, median and mode. But the average might obscure information. Treating every customer as the average might be disappointing for your fan, too little too late for potential defectors and irrelevant for the rest. In order to handle this, you need to introduce segmentation.
Presentation is important in data handling. Dashboard is a living report of what’s going on in the business and they can be powerful tools for decision makers, since they can create good data habits. The biggest failure of dashboards is that they try to do too much. Good dashboard collect data on one screen, they need to be updated regularly and they need to emphasize important issues. Every reported statistic should give insight where you are and identify if potential action is needed. Trackers are another important tool for collecting data. The idea is to sample same basic data over and over again, at regular time intervals, so that you can follow a trend. Some of the questions we can follow with trackers are: competitive benchmarking, customer relationships and experience-triggered measurements.
When we talk about errors in data, we should look from two perspectives: not only accuracy but, but also time perspective. It is not possible to measure without error, so you need to decide the value of waiting until you have better data, or of putting in more time to do more accurate calculation.
You don’t know the future, but it’s helpful to have a go at predicting it anyway. Tom Peters says: “Some achieve a reputation for great success when in fact all they have done is take a chance that reasonable people wouldn’t take.”
Good managers have bias for action. This could be a problem, since this means that they don’t take time for analysis. Predictive analytics (the big data of forecast) is powerful but needs investment and lots of data. Trend analysis are weak at predictive features since they are following a path of what happened and the most important guidance is often the prediction that something is about to change. Having strong forecasting team can help you with forward guidance. This allows you to set direction for the business. In forecasting the reporting and understanding of errors and uncertainty is vital.
The psychologist Gary Klein has created a technique to improve the quality of the analysis that he calls a “premortem”. This are actually lessons-learned session that looks at what went wrong, but done up front before projects. Modeling the risk into your decision is extremely complex. But any forecasting has some level of uncertainty. So, you need to evaluate a risk. If you don’t, you are not making decisions, you are hoping for the best. When you look at the potential for returns in business, you need to also connect them with risk profile. You need to decide about your risk appetite. The expected return is always the sum of all possible returns, multiplied by the probability of achieving that return. And probability is in this case risk factor. The most important forecast is return on investment. One of element of calculating returns is using NPV – net present value. NPV estimates the lifetime value of an investment by adding and subtracting all the cash flows related to it, each year, for the duration of investment.
Just because the data tell a story doesn’t mean the story is always true. Potential cooking of numbers can be done by: selective evidence, bad causation, a weird time frame. Data cannot usually tell you why, only what.
When there’s a lot of data, how can we avoid paralysis by analysis? Some potential action can consist of:
• When it is sure thing, do it.
• Don’t bother if cost of getting information is higher than its value.
• Eliminate options by comparison.
Learning how to compile and analyze data is only half the battle. Learning how to use the data to clarify debates, rather than complicate them, is a complementary skill that’s possibly more valuable.
Rule one of being a data-driven manager is to trust your own information. In today’s world success gives you authority of knowing. But success can be also result not of careful planning but more of luck. Daniel Kahneman points: “A stupid decision that works out well becomes a brilliant decision in hindsight…If you group successes together and look for what makes them similar, the only real answer will be luck.”
While many business decisions could benefit from more data, few benefits from data alone. Instinct is a useful decision-making tool for survival that has developed over thousands of years. The important thing in business is to use it when it gives you the best information you can get. Time is always a factor in using data for decision-making. Relevant data is not always available. Experiments are also one tool to help with data usage. In Google they run 1000 experiment at any point of time.
We often use outcomes to modify our prior beliefs without justification. This is an aspect of what’s known as Bayesian statistics. When we talk about conditional probability the chance of A, given B is not the same as B given A. Giving your team a second chance may be a good management. In risky or uncertain environments, a one-of results often doesn’t mean what you think it means. Many business processes are set up so that they protect status quo. When we look at budgeting we actually take starting point last year performance. Instead we should break activities inside business models and estimate what costs they should take, so that they can be profitable and efficient. Zero- base budgeting is concept from Peter Pyhrr.
Using data to improve customer expirience is new sales approach. Since so far sales approach customer more from company profitability and not value for the customer. This can be seen in account-based marketing (ABM). ABM uses data to put both right: have limited number of interventions, with the correct people at the right time.
Agile methodology is another attempt to bring company closer to customer. Agile methodologies break the project into smaller »chunks« which are worked on by small groups, delivered in days rather than months. Management is bottom-up and feedback driven, rather than top-down and ideological. Agile methodologies, driven by constant feedback, is a perfect example of building data-driven decisions into an area normally governed by rigidity or management by instinct.
If you ignore data and trust your gut for long enough, you might do well in the short run. But you’re not likely to have a hot streak that’s going to last your entire career, especially when you’re up against the combined power of instinct plus data in your competitors. Gut feel can be flawed because:
• Gut feel ignores other people’s need
• Gut feel is about the past
• Gut feel success or luck
• Gut feel creates groupthinking
• Gut fell stops us learning
The number one gut feel bias is an illusion of superiority.