What does it mean to be a data-driven company?
A data-driven company has a culture where the data is respected and adequately utilized to make business decisions across an organization. It means that each business decision, minor or significant, must be justified based on numbers/metrics.
This process demands different steps; first, you must collect the data, which implies that you have control and track of what happens in your product. It means not only the final number (like, how many ‘x’ products you sell) but also the user’s entire behavior, such as clicks, pages visited, and so on. Secondly, you must normalize the data, which means cleaning the data, for example, removing redundant details that you do not need to process or doing minor modifications on the fields to unify the values. The last step is analyzing and showing the information. Perhaps the most difficult one, given that it implies the challenge of knowing how to interpret your data to present it. This approach permits avoiding bias for personal hypotheses and insolation resolutions. Of course, it can not promise that it will always make the best decision, but it is a powerful tool to increase the guarantee of success.
According to a new research related to the value of becoming a Data-Driven business (Ejofodomi, 2022), there are other benefits, such as helping you outperform your competition, improving customer retention, and identifying your business needs, just to name the most relevant ones. In this context, we are talking not only at the management and product level but also at the entire company, from the development, sales, and human resources departments.
In this post, we will discuss data-driven from the engineering perspective, the benefits of applying it in your company, and how this can improve your daily life.
Learning by example
Let’s take as an example a company that generates a two-sided marketplace that connects companies with qualified professionals. When new tech candidates visit the website, they have to follow an onboarding process with multiple steps including complete different information related to their career, education, work history, skills, etc.
On the other side, the marketing team tracks the conversion numbers daily. So, for example, let’s say that on Oct 4, we got 81.57% of the new candidates to complete the onboarding and fill in all the required data. So, the percentage of new users finishing the onboarding process and completing the steps is 81.57%. This number is related mainly to our ‘Conversion Cost’; in raw words, how much money the company invests (or it costs) in Ads, Marketing, events, and so on.
Example of an onboarding Conversions simulation for last weekWhen a new feature is requested
Let’s say, for example, the product team requested a new feature that makes it mandatory to update a resume. It’d impact the platform in two ways. On one side, the fact that each candidate has a resume uploaded will make them more complete, but on the other side, our conversion numbers could decrease.
Tracking and tracing
Behind the scenes, we might track what the users do all the time. So, for example, at the specific step described before, in which the user must decide whether to upload the resume, we keep tracking in our database what their decision was. But we do not track only the specific decision; we also add extra information related to the user’s context, such as his skills, years of experience, which features have been enabled, etc. Finally, we might experiment with it activating only 50%, with the objective of, after doing analytics, comparing results and taking action on it.
Analytics
We might sync all data with a big data analytics software like Looker, which permits easy exploration, analysis, and sharing of real-time information. This will give us complete control over what is happening on the platform. With all the data synced, we might create specific graphs (using SQL) to track the conversions. If we activate only 50% of the cases it’d mean that half of the users will have to upload their resume.
Submit profile, control vs. with resume.
The image above shows how the new request affects the conversion rate (the user must upload the resume). In general, we lost (in yellow) around 10%-20% of conversions. But, on the other hand, we have Candidates with better completeness profiles, which can impact more hires. This also shows that you shouldn’t make decisions based on this graph alone. Instead, you must consider and complement it with the company’s vision and its targets. For example, the company might evaluate if the preference is to have less active candidates with more quality profiles or prioritize quantity over quality.
Conclusions
Nowadays, it is a must to work using data analytics to feel comfortable and, most importantly, to have complete control of each contribution to change in the app.
Every time you deploy a new feature, you will have enough business reports to know that if a problem appears, you will likely catch it sooner; As a consequence, it will save you lots of money.
Working with this kind of structure will make your day more enjoyable and relaxing, given that you have less stuff to worry about and focus on the business when you are coding; in the end, that is all that matters.
References:
Business Intelligence (BI) & Data Analytics Platform. (s. f.). Looker. Recuperado 18 de octubre de 2022, de https://www.looker.com/
Ejofodomi, E. (2022, 7 enero). What’s The Value Of Becoming A Data-Driven Business? Forbes. Recuperado 18 de octubre de 2022, de https://www.forbes.com/sites/forbesbusinessdevelopmentcouncil/2022/01/07/whats-the-value-of-becoming-a-data-driven-business/?sh=4167ded43776
Hodgson, P. (s. f.). Feature Toggles (aka Feature Flags). martinfowler.com. Recuperado 18 de octubre de 2022, de https://martinfowler.com/articles/feature-toggles.html