Increasingly, decisions in companies have to be based on evidence and less on managers' instincts or intuitions. Of course, each manager's personal experience and knowledge will always influence their decisions, but essentially, facts must be the basis for decisions. And where do these facts come from? Facts, and the knowledge that supports them, result mainly from the data that organizations have and how they organize them , that is, how they transform data into information and subsequently, into knowledge and wisdom.

However, if the data has what is technically called a lack of “quality”, decisions that may arise from making decisions based on it may turn out to be incorrect. So, what should quality data sources look like? Let's see, they should be:

1. Exact - The sample used must reflect the population.

For example, to study price evolution you cannot simply study a small sample of hotel data or a few hotels in a chain. All available data or a statistically valid sample must be used.

2. Reliable - Data does not change if extracted over multiple periods.

For example, if we are going to analyze reserves for a certain period of time that has already passed, and take two samples on different dates, will the data contained in both samples undergo changes? They shouldn't suffer.

3. Accurate - It must contain the information necessary for the problem under study in the most atomic form possible.

For example, if the problem is to study reservation cancellation patterns by nationality, the data must be available by reservation and not aggregated by segment, distribution channel or other attribute.

4. Impartial - The extracted data must not be affected by any criteria that could call into question its statistical validity.

For example, if, to predict the average future price, the analyst does not include in the data sample the reservation data of a specific operator or nationality, this will influence the models created.

5. Valid - It must be ensured that adequate processes have been implemented for data extraction, organization and analysis.

For example, if data was extracted that says that in a given year there were a certain number of room-nights occupied, that number must be exactly the same as what the hotel's PMS indicates for that same measure.

6. Appropriate - The data to be used must be appropriate to the problem under study.

For example, if the objective is to study no-shows at a hotel and the data in question does not allow distinguishing no-shows from cancellations, the data source is not appropriate for the problem under study.

7. Timely - The data must be related to the period under study and must contain all possible observations for that period.

For example, if you want to develop an overbooking model for the next 5 days, but the data source used is only updated once a week, it should not be used.

That said, the reality is that when worked well, data can give us highly reliable answers to the questions that generate the results we are looking for. However, it is imperative to pay attention to their quality so that these answers are also the most accurate.

‍