We search for complex connections between state-of-the-art technologies or examine the mesmerizing base of the latest technique. Yet, data science or  AI is not only bragging about new exciting methods that boost accuracy by 2% (which is a significant gain) but about getting technology and data work for you. It will help you maximize sales, understand your clients, foretell future errors in process lines, or create an insightful presentation, submit a term project or have a good time with your friends working on a new idea that will advance the world. And in this sense, all can — and to some amount should — become a data scientist.

We previously discussed what makes a great data scientist and what you should learn ere you set to a real project. In this post, we'll discuss the process of creating a backbone data project in easy steps.

You have an excellent opinion in your head

the one you have liked since you were a child about becoming a toys-cleaning android or the one that just came into your mind about accessing the clients in your shop by assigning them fortune cookies with foresight based on their shopping preferences. But, to make your plan work, you want the attention of others. Find a compelling story for it; make sure that it has a latch or a captivating purpose, that it is up-to-date and appropriate. Searching the narrative structure will help you decide whether you have a story to tell.

Such a story will be the basis for your business model. Ask yourself: What is it that you acquire, what sources do you want, and what value do you give to the client? For what costs are consumers going to spend?

The best way to do this is the business model canvas. It's cheap and straightforward; you can create it on a sheet of paper.

The initial practical step is getting data to fuel your project. Depending on your goals and field, you can seek for ready datasets available on the Internet, so for example, this combination. You can decide to scrape data from websites or access data from social networks by public APIs. For the new option, you want to write a small program that can download data from social networks in a programming language; you feel the most comfortable. For the cloud choice, you can spin up a single AWS EC2 Linux instance (nano or micro), and run your software.

The best method to store the data is to use a single .csv format with each line, including the text and metadata, such as the person, replies, timestamp, and likes.

As the amount of data required, The rule of thumb is to get more data as possible in a reasonable time, for example, a few days of operating your program. Another essential consideration is to collect as much information as the machine you are using for analytics can manage. How much data to get is not a precise science, but it rather depends on the technical restrictions and the question you have.

Ultimately, in managing and collecting data, it is crucial to be devoid of bias and do not be selective about the inclusion or exclusion of data. This selectivity involves using discrete values when the information is continuous; how you deal with needs, outlier and out of range values; arbitrary temporal ranges; capped values, volumes, scales, and intervals. Even if it is arguing about influencing, it should be based upon what the data says–not what you want it to say.

To perform an accurate analysis, you want to obtain the proper tools. After accepting the data, you need to select the appropriate device to explore it. To make a decision, you can write down a list of analytics features you think you require and compare possible tools. Seldom can you use user-friendly graphical tools like Orange, Rapid Miner or Knime? In other circumstances, you'll have to write the analysis on your own with such languages as R or Python.

With the tools and data available, you can explain your theory. In Data Science, techniques are statements of how the world should be or is and are derived from axioms that are assumptions about the world, or precedent theories (Das, 2013). Models are implementations of the approach; in data science, they are often algorithms based on assumptions that are run on data.

To evaluate your theory at a first step, in line with the more general and traditional content analysis, you can pinpoint inclinations present in the data. One method we use quite a lot is to collect significant events that have been listed. Then you can try to design an analytics process that gets these trends. If analytics can see the patterns you specified, then you are on the accurate track. Look for cases where analytics finds the latest trends. Confirm these trends, for example, by exploring the Internet. The outcomes are not going to be reliable 100% of the time, so you'll need to decide how many falsely reported trends (the error rate) you want to tolerate.

When you have your business model and proven theory, it is time to build the first version of your product, the so-called minimum viable product (MVP). It can be the beginning version that you give to clients. As a minimum viable product (MVP) is an output with just sufficient features to provide early clients and to provide feedback for future development, it should focus only on the core functionality without any fancy solutions. It would be best if you stuck to pure functions that will work at the start and expand your system next.

In principle, your center should be on the future development of your product, not on system operation. For this, you require to automate as much as feasible: uploading to S3, beginning the analysis or data storing. In this report, we examined automation in detail.

The other face of automation is logging. When everything is automated, you can observe that you are losing command over your system and do not know how it works. Besides, you want to know what to develop alongside, both in terms of new features and solving problems. For this, you want to set up a system for logging, monitoring, and measuring all essential data. For instance, you have to log statistics for the download of your information or upload it to S3, the time of the analytics process and the users' behavior.

There are several tools to help you log server statistics like RAM, CPU,  network, code-level performance, and error monitoring, many of them having a user-friendly interface.

You probably know that Machine Learning, AI,  Data Science, and other latest developments are all about reiteration and fine-tuning. So, when you have your MVP automation, running and monitoring in place, you can start improving your system. It is time to get freed of defects, enhance the overall performance and stability, and add unique functions. Executing new features will also allow you to offer new services or products.

Finally, when your product is active, you need to present it to the clients. It is where your story behind the data and business model comes to help.

First of all, study your aim audience. Who are your clients, and how are you going to sell your product to clients? What does the audience you are dealing with giving your merchandise to know about the subject? The story requires to be framed around the level of information the audience already has, correct and incorrect:

Novice: first exposure to the issue, but doesn't want simplicity

Generalist: aware of the question, but looking for a summary understanding and significant themes

Managerial: in-depth, actionable understanding of complexities and interrelationships with access to the article

Expert: more discovery and exploration and less storytelling with enough detail

Executive: only has time to learn the significance and conclusions of weighted expectations

Afterward, visualize your data and incorporate trends, importance, and proportion you built your project into a narrative. Your story about the product should never end with a fixed event, but preferably a set of questions or options to trigger an action from the audience. Never forget that the goal of data storytelling is to strengthen and energize critical reasoning for business decisions or to buying your product.