The digital specialist of our agency, Taras Humeniuk, has investigated how the accuracy of data and the effectiveness of management decisions in business are interconnected.
What decisions can be considered pondered?
Before crossing the road, you should look around. To the left, to the right, and forward. The amount of data for analysis is small, and the criteria are obvious. In e-commerce, the number of data and variables increases significantly: we have indicators of reach, average session time, interactions with the product card, and purchase initiation. And also we have the purchase itself, ROI, ROMI, ROAS, LTV and a number of indicators, each of which can affect the decision.
For example, let’s consider a simple structure of an advertising campaign with branded (No. 1) and other (No. 2) search queries. The purchase price in campaign No. 1 is 15 UAH, in campaign No. 2 is 32 UAH. The average price of the product is UAH 70, the margin is 30%.
Based on this, campaign No. 2 is unprofitable. It needs to be stopped and the budget transferred. Based on this data, we can make a decision, let’s call it “Disabling Campaign No. 2”.
Now let’s add more data – the average order check. In both campaigns, it is the same – UAH 210. That is, with a margin of 30%, campaign No. 2 is still profitable (210*0.3=63). And the decision “Disabling campaign No. 2” will not be appropriate. Without data on the size of the average order check, there was a risk of making a decision that would lead to the loss of one of the revenue channels.
Therefore, before disabling campaigns, transferring budget, or before making any decision at all, take into account as much data as possible.
The fewer unknowns, the more pondered the decision can be called.
What data is accurate?
Data in which the quantity and quality of errors are reduced to a minimum can be considered accurate in digital marketing. We will discuss the issues of data collection and attribution below.
Data collection problem
After the iOS 14 updates, users got the option to allow or not allow the transfer of information to Facebook. Therefore, some data may be lost. This is not critical for the advertising cabinet of some, for example, local media. But if we are talking about an IT studio that develops applications exclusively for iOS, then in case of inactivity, approximately 80% of the user behavior data may not be obtained.
Here we are talking about both the model and the window. Each attribution model has its pros and cons. The last click model is often used. It gives an easy-to-understand picture. But if several channels of paid traffic work at the same time, it becomes more difficult to understand the role of each in a specific conversion. And a simple solution will not be the best option here. If to talk about the attribution window, then when choosing or not choosing it, it is advisable to understand how much time the client needs to make a decision. Obviously, the time frame for analysis, selection and purchase of a desk, a garage and a notebook will vary. But with proper work with the buyer’s path, the false influence of the attribution factor on the decision-making process can be significantly reduced.
Which representations are correct?
After collecting and segmenting the data, the question of convenience of perception and understanding of the relationships between all indicators appears. What data is better to present in a table? Which ones are a pie chart, and which ones are a tree map?
This is most evident when presenting data analytics to a client. Therefore, these data should be presented in such a way that even a child can understand what the slide is about without explanation.
Considering the trend-need of business in digitalization, the correct work with data can give a competitive advantage in the fight for the attention of the consumer. That is why, master new technologies, learn to make accurate, balanced and effective decisions.