Data interpretation

Write Close
Do you have any questions? Contact us!
I agree the Terms of Service
published July 5, 2018

I would like to share with you the theme of data interpretation. For many years I had to work actively with data. I met different cases during writing reports for users and top management, analysis of requests for reports, and of course analyzing the requirements of the customer.

I often see that analytics make unambiguous conclusions, but they are quite controversial. Or, when composing a request for analysis, they try to consider the data only from one point of view. That's why I would suggest you to look at your query at some other angle and find other conclusions.
How does your management see your data?

Some digits and rectangles of different sizes. Without any comments and explanation it is an utterly unnecessary picture. The main conclusion which an analyst likes to make: if something grows, it means it's okay, if it falls, then it's terrible. Analysts often choose for the bosses only that increasing data trends without any explanation. And what has grown? Why? What does it mean? If an analyst does not have answers to these questions, then this is a mistake.

There are many examples where the trend is growing, and this is bad. For instance, if we consider the situation that the trend for the number of appeals or complaints is increasing, then we are working poorly and something is going wrong. If the period of requirements gathering and clarification from a customer is growing, in this case, development stands still.

Making decisions

It is necessary not only to receive data but also to make decisions based on the information received. Otherwise, it is an analysis for analytics, i.e., a waste of time. When I processed requests for various reports in telecommunications company, I usually asked for what purposes you need this data. First, it helped to weed out unnecessary queries — i.e., if a person does this just because someone told him, and he does not understand
or is not interested in the final result, then I would not want to spend my time on it.

Secondly, a person can miss an indicator, watching the situation on his part and this will not give him a full picture of what is happening. For example, if an analyst has the primary task to check the efficiency of the marketing campaign, then most likely he will ask only the data relating to the period of the action and to the subscribers who fall under it directly. But he may not think that at the same time a discount was launched, and his marketing campaign does not bring enough revenue, because some subscribers are served at a reduced price. Finally, the result will be misapprehended, it will be impossible to draw conclusions and take some corrective actions.

Data forecasting

An excellent skill is data prediction. If you already have some quarterly / monthly or weekly data, then, based on the average, you can predict how the system, subscribers or revenue will behave.

A good example is the evaluation of labor costs for the customer, especially in the commercial version. For example, we have a project with one developer; its price is $25 per hour, the duration of the work is 6 hours or 150 dollars a day. Everyday we ask customer three new questions and get only one answer. Each answer is one hour of developer work. By the fifth day, we have eight unanswered questions, or 200 dollars, wasted for "the work is in place."

In total, we can say that developer works 4 hours a day and sits idle for 2 hours. If this continues, then for a period of 30 days we will have a situation that will accumulate 48 unanswered questions (30 * 8/5). The customer will pay totally $ 4,500 (30 days 6 hours of work at $ 25 per hour) of which $ 3,300 for useful work and $ 1200 for idle. This example, of course, is very crude, but it perfectly tells the customer how much his silence costs. There are, of course, some clients who are satisfied with it.

Tracking Results

If you received the data and having drawn conclusions, it is necessary to take some further action. And after that, you must always track the results of your manipulations. For example, you have an inefficient site; it does not bring you the right amount of orders. You decided to spend a budget of $ 1000 to set up your site's SEO. After a while, the management asks what the result is

- It seems site becomes work better. We now like the site more.

And how much is it in percentage, in numbers? Due to what? What did you do? What actions have been taken? Without specific figures and analysis, this does not provide answers to all the questions posed.

There also might be a problem with the executor. You convinced superiors to give you a budget for optimizing SEO, but did not look at the results, or looked, but did not make any conclusions. And most importantly, they did not bother to tell the result of the analysis of the effectiveness for their superiors, and next time you are unlikely to have approval on this application. Not only that the leadership does not understand what is SEO, but there is no efficiency from it. Why should we spend any money on it?

Who needs information and what are the goals of this?

The first question that you should ask yourself when preparing the analysis is who needs the analytical data that I will provide? Often, top management does not require data on the revenue of one subscriber or data on the inefficiency of one employee, unless the task was about it. The management wants to see trends, comparison for the period, conclusions and specifics.

If you need to familiarize the new employee with the work process, then you should show him all data in the details. For example, you may reveal the sample of data from data sources, then turning it into accumulative indicators, compiling a report, and describe components of each indicator.

Perhaps you provide data for an employee who works with complaints. Data showing that there is incorrect traffic charging for a group of subscribers. Then you have to understand the details of charging on the example of one subscriber. To go through the whole process, and to see all the services, traffic, charges and so on.

The second question that you need to ask yourself, what conclusions does analyst have to make based on your data?

As a rule, if I come to the management with some complaint, and do not provide him with solutions, then this looks like a general discontent. If you show figures, then the management should understand why they need it, what follows from this and hear your suggestions on this matter.


1) Harvest of garlic has grown on the outskirts of city A. So what?

2) Harvest of garlic has grown on the outskirts of city A. They are the leading suppliers of products to the city B. Last year there was a shortage of garlic, so we decided to earn on it and exported 300 tons of garlic there. And because of city A has a high yield this year, and starts to dumping prices drastically, our income from the exported product may fall sharply. Therefore, we propose to distribute a part of the exported product to city C, in which there are no products from city A, this will help us … and so on.

Problems of interaction

This is a common problem for large companies. For example, I analyze the data to decide on the launch of a new marketing campaign. I run it and switch to the next run analysis. At the same time, I spend corporate money on advertising and employees time to create and process it. But the review of the efficiency of the campaign is handled by another department, which does not influence me. Time after time they see the analysis results displaying ineffective campaigns or promotions, but do not sound alarm, do not give recommendations and do not ask management to take decisions. Chariot turns, work is done, efficiency is zero.

Strange data

If you received data that shows you a big splash, and most importantly, you exactly got the result you were expecting, do not rush to run with this news to the leadership. I would suggest that you look at this data for a period, compare and thoroughly dig into the causes.

For example, you launched a new tariff plan and predicted that the efficiency of it will be revenue of $ 500 per week. We decided to wait two-three weeks and after that draw conclusions. You request revenue in 3 weeks and see that the first week revenue is $ 50, the second is $ 60, and the third is $ 800. I knew it! — you think at this moment. If I were in your place, I would begin to doubt the correctness of this data. Perhaps there was a failure in the billing system? Maybe the one who collected the information for you, today did not get enough sleep and put an extra zero by mistake. And if all the indicators are indeed correct, then it is necessary to find the reason for such growth. There is a high probability that there are some factors that caused this splash. Then you can safely go to the managers and say — the revenue from this tariff is $ 300 per week, but this figure cannot be considered fair, because at the moment there is a football championship in the country and many tourists arrived in a couple of days and use SIM card with this tariff plan. And if you exclude this week from the analysis, then the average revenue will be $ 55 per week.

Ambiguous indicators

There are cases when it is necessary to make an assessment, but there is no exact method from calculation and analysis. For example, we do not have single-valued KPIs for calculating employee performance in a sprint. One developer can close 10 tickets in 40 hours, another in 45 hours will close only three. But the problem consists not only of developers work speed. Here it is necessary to take into account the complexity and volume of tickets. In such situations, you need to calculate the average performance of the entire team and use it to conclude the efficiency of each employee. Or, take the average figure of a particular employee and see if this is his standard time of work. You can not conclude with only time or quantity; you need to be more flexible and see all the factors.

For example, we're analyzing sprint planning results. What may be easier than data grouping from each developer out of the columns planned and spent? We compare data, the plan and fact practically coincide in each sprint, but suddenly, in one of the sprints, we suddenly see a big difference between them. Does it mean that we planned sprint poorly?
It is necessary to understand more carefully the reasons that led to such a gap. In the sprint 19 (on the picture) we attracted interns to the project, and the developers began to work several times slower distracted by them, while the number of hours spent increased. This indicator is typical under the existing conditions.

In conclusion, I would like to offer you, even though you

trust yourself more than anyone, ask another person to look at your analysis result. Especially if you are just a beginner analyst. It is always useful to get questions about data that you did not foresee, and they can lead you to a different point of view. Look at your analysis with a different look, do not make hasty conclusions, consider different versions.
Did you like this article?
Share article on social networks
Worked on the article:
Juliana Amelina
Deputy Chief Executive Officer
Made on