What Are the Expected Results of a Data Science Project?

Daniel Morales
Jun 11, 2021

From a company perspective, data science projects should always be viewed as experiments. Remember that we are talking about science, and science bases many of its theories on the results of a series of experiments. From here, many companies start with the wrong assumptions, thinking that the results are exact sciences, of which there must be a single, true answer. 

The reality is that many data science projects fail because of the lack of iteration that is needed once the first results are obtained, and because they do not adopt a scientific approach to the process, that is, an experimental approach. But what are the expected results? Well, the most common results expected from a data science project are:

  1. APIs
  2. Integrations
  3. Applications and/or Platforms
  4. Reports
  5. Presentations

1- APIs

APIs are a set of subroutines, functions and procedures (or methods, in object-oriented programming) provided by a certain library to be used by other software as an abstraction layer. 

This sounds a bit confusing, but it's easier than it sounds. This is code that can be shared between machines. Here the requests are not made by a user from a browser, but by one program to another program, and the result is chunks of code that can then be read and processed. 

APIs is the most common form of expected outcome for a company that intends to develop a data science project, specifically talking about predictive models, as it allows to easily integrate the solution within the company's current internal programs, without the need to worry about deep integration, or the incompatibility of programming languages, devices or internal systems. 

This is how a company that has a strong web, IoT, or mobile presence can immediately use a data science solution without the need to invest large resources. 

The requirements for an API, as an expected outcome of a data science experiment are:
  • Well-written, understandable and reproducible help pages or documentation
  • The code must be well documented
  • The code must be version-controlled

2- Integrations

Integrations are a bit more complex from a technical point of view, since they involve integrating a solution within the company's current systems. 

So, for example, if all the current web development was done in Java, and the machine learning solution was delivered in Python, the software engineering team will have to figure out how to integrate both languages into the stack, either through internal microservices or other application coupling techniques. Let alone if the company has a monolithic stack. If the latter is the case, you should opt for an API. 

3- Applications and Platforms

Another common solution is to create an external service, completely different from the company's main service. Here the aim is to host the service in a different domain, where the objective is to access only to obtain the results of the predictive modeling. This is common for companies that use engineering as marketing, making predictive programs to add leads to their pipeline of prospects. 

The requirements for these applications and web pages, as an expected outcome of a data science experiment are:
  • Ease of use of the tool
  • Help pages or documentation
  • The code must be well documented
  • The code must be version-controlled

4- Reports

Here we would no longer be talking about predictive models, but about data analysis in general. The most common and expected result is a report or series of reports where it is expected to understand the historical data, the reason for the historical results and the conclusions of the same. 

They are usually full of statistical data useful for decision making. They are useful for management, marketing or human resources committees. There are many formats for this, but the ideal would be not only to present the report, but to have the opportunity to make a presentation and tell a story about the data. 

Ideally, the reports provided to you should be
  • Clearly written
  • Include a narrative around the data.
  • Creation of an analytical dataset
  • Analysis
  • Clear and even interactive graphics
  • Concise conclusions
  • Omitting unnecessary details
  • Reproducible

5- Presentations

Presentations are where data scientists tell stories with data. A detailed but conclusive report on historical data is expected for decision making. This helps any area of the company, and can be presented at any business committee. 

The exact same criteria for presentations:
  • Clarity:
  • Include a narrative around the data.
  • Creation of an analytical dataset
  • Analysis
  • Concise conclusions
  • Clear and even interactive graphics
  • Omitting unnecessary details
  • Reproducible


As we can see there are several types of results when we expect to run a data science project. We see the importance of having clear objectives, what we can achieve, but most important of all: take the project as an experiment, not as a magic solution to all our problems. 

“What Are the Expected Results of a Data Science Project?”
– Daniel Morales twitter social icon Tweet

Share this article:


Post a comment
Log In to Comment

Related Stories

Jan 06, 2023

Top 2 Online Data Science Courses to Improve your Career in 2023

The discipline of Data Science is expanding quickly and has enormous promise. It is used in various sectors, including manufacturing, retail, healt...

By Nikos_DataSource
Nov 25, 2021

5 Tips To Ace Your Job Interview For A Data Scientist Opening

5 Tips To Ace Your Job Interview For A Data Scientist Opening.PNG 795.94 KBImage SourceAspiring data scientists have a bright future ahead of them....

Daniel Morales
By Daniel Morales
Nov 12, 2021

When to Avoid Deep Learning

IntroductionThis article is intended for data scientists who may consider using deep learning algorithms, and want to know more about the cons of i...

Matt Przybyla
By Matt Przybyla

Join our private community in Slack

Keep up to date by participating in our global community of data scientists and AI enthusiasts. We discuss the latest developments in data science competitions, new techniques for solving complex challenges, AI and machine learning models, and much more!

We'll send you an invitational link to your email immediatly.
arrow-up icon