The Data Product Life Cycle Part 1 – Data Analytics Consulting

The Data Product Life Cycle Part 1 – Data Analytics Consulting

March 4, 2021 big data consulting Data Driven Culture data science Data Science Consulting 0
data engineering consulting

Photo by NEW DATA SERVICES on Unsplash

Many companies treat data as an afterthought, like a chef throwing parsley on a dish and calling it a garnish.

That’s because many companies don’t live and die by their data. They don’t eat, breathe and sleep data.

Well, I have.

Or at least, I have worked for companies whose entire bottom line relied on their ability to manage, analyze and visualize data.

These companies have created standardized metrics that are often easy to explain as well as data visualization and actionable reports.

In these companies, developing a dashboard needed an entire process. There was a ritual, a clear set of steps, to go from idea to product.

And these companies, these companies were quite profitable although they are somewhat anonymous.

Many might wonder. Why can’t the companies just build their dashboards with their metrics? The problem lies in the overall process. Many companies don’t have a good process for developing data products and instead build a dashboard or machine learning models that don’t fit in their overall strategy.

At a data product company, you have to think bigger. You can’t just throw a quick dashboard together at these companies and hope no one asks about the specifics of each data point.

You also can’t build a dashboard or model and be ok with a few weeks later, no one ever looking at it again.

There are plenty of BI and data engineers that build dashboards. Many companies will have well over one-thousand dashboards that often lie neglected in a dashboard graveyard of forgotten metrics.

You can’t let that happen when your bottom line is dependent on the usage of your dashboard.

Building data products is a great way to build revenue.

However, it does take a lot more work than most people think. A single dashboard could take 3–6 months at least to build.

During that time you will spend time researching, parsing, QAing, and designing a dashboard that users want to use. These dashboards are treated like what they are.

Software.

Not some BI afterthought. But a product that requires version control, QA, product management, and DevOps.

Data products can be very profitable, as long as they are treated justly.

So how do you develop a data product?

How To Develop A Data Product

Building a data product.

A data product that truly drives impact.

Doesn’t happen overnight. One issue I have seen is that companies, business owners, and executives can get so excited by the possibilities of what data can do, that they don’t have a clear process for getting to that end state.

Although I do find overly rigorous processes restricting, having a clear set of guidelines will ensure that you increase your success rate in the data products you do develop.

Below we will provide some tips for your team when you’re developing your data products.

Define A Business Problem

Creating successful data products like dashboards, data APIs and algorithms first requires that you have a clear business problem you are trying to solve.

Building data products that don’t tie to a business goal or problem will lead to an unsuccessful product. Even if it works as intended. This is because regardless of how accurate the algorithms you build are or how aesthetically appealing your dashboard looks.

Without a clear business problem or goal, it’s an orphan tool. It’s difficult to apply the insights to a tool that doesn’t have some clear business problem.

IN turn, this is why before spending months on building a data product, there should be some formal set of business goals written out. This may feel like a hindrance and merely mundane bureaucracy that provides no real progress in terms of the product. But I have seen and been part of a fair share of projects that were rudderless. There was no clear goal or vision from the start and although there might be a lot of initial excitement, due to the tools being used or an outcome that sounds impactful but doesn’t align with the companies currency resources or plan.

Analyze What Data You Have

This step has two distinct parts.

First, defining what data your team currently has access to. Do you have data from Facebook Insights, Zendesk, Workday, Salesforce, or some other software?

Maybe you have internal tools.

Or maybe there are public data sources your team can access.

Taking a general inventory of your data sources, or what data sources you could have access is a vital step because your team might discover that it can’t build the data products you had hoped or you might discover new insights that you can drive.

Once you have listed out the data you have access to that would be relevant to your project, then you can start to analyze it for any easy insights or trends.

This step doesn’t need to take long. The goal here is to understand your data and its various nuances.

During this step in your process, you can start to write up some general notes about trends, charts, and rough metrics you notice.

You might even be inspired for possible new projects in the future.

Define Your Final Product

Defining your final product, what it will do, and generally how it will look is a necessary step before setting up a plan.

I have seen several failed projects occur because the final product wasn’t clearly defined

To contrast that, I have seen multiple successful projects where the team spent an extra few days defining the final product.

What will it generally look like?

How will users interact with it?

Will it be an API, a dashboard, a model, or a report?

Several problems occur when your final product doesn’t have some form of definition.

First, if there is not a clearly defined plan, projects can spiral for weeks without making any real progress. Generally, this is because there tends to be a lack of clear direction which infiltrates every team slowly as it becomes clear that there is no plan.

Second, scope creep becomes much more likely. When a project isn’t clearly defined in the beginning, the product can quickly change into a different final product.

Both of these paths can quickly lead a project to failure.

Develop A Clear Project Plan

With a clear definition of what your project will look like when it is done, developing a project plan becomes simpler.

A project plan should take into account both the technical and business tasks that will need to be accomplished for the project to be successful.

For example, if your project also needs marketing, PR, and business development, then these steps should also be accounted for.

This is because sometimes some of these tasks need to happen concurrently with the development process so you will need to know where they overlap.

When a project is well planned, even if you’re just using a Google Sheet to track the tasks that need to be completed, it will succeed.

Tools like Asana and Kanban boards aren’t useful if you don’t spend a good deal of time planning what needs to be done.

Types Of Data Products

Data products manifest themselves in multiple different ways. You can develop and output dashboards, APIs, algorithms, reports, and more.

Each of these products has different paths forward and we wanted to provide a few insights about developing each of these various products.

Dashboards And Metrics Based Tools

Dashboards. Arguably one of the easiest data products to develop. With all the low-code options like Tableau and Looker, quickly prototyping and then putting out a dashboard doesn’t take much.

But.

There are still a lot of underlying concepts you need to consider.

Like developing or picking the right metrics that will help align your dashboard to your business needs. There are hundreds, thousands of metrics to pick from across multiple industries.

So step one, is to decide what you want the dashboard to drive.

Are you trying to increase customer satisfaction, reduce fraud, reduce the number of days to close a deal, etc? There are so many choices to choose from when it comes to goals.

This is why you need to pick one when developing a dashboard and or the metrics for the said dashboard.

Dashboards that are just an amalgamation of multiple metrics that don’t provide a coherent picture of what is currently happening tend to fail.

Users need to be able to look at a dashboard and understand how their metrics reflect their real-life business so they know what to change.

When dashboards provide these clear metrics, then business executives can act with conviction. They can make strategies that align with their business and their business needs.

Leading to a possible competitive advantage.

Algorithms And Models

Algorithms and models are another common final data product.

These are far from simple data products.

Whereas in dashboards, where metrics and charts are usually pretty self-explanatory.

Algorithms and models have a lot of risks that pose challenges when it comes to building tools.

However, when done well, these models can provide enormous benefits.

Think about Amazon’s ability to recommend products or BeyondPricing’s ability to recommend pricing for your Airbnb

These models increase the profits these companies can make.

Unlike dashboards and metrics that often require human interactions for there to be some form of cost decrease or income increase. Models and algorithms can be implemented in such a way that they automatically provide their data outputs in a way that leads to a better bottom line.

Often, these models and algorithms allow companies to scale without hiring more employees.

Instead, these machine learning models can help better allocate resources, calculate fair pricing, and optimize company processes. All with minimal monitoring.

The trick here is, developing machine learning models that are robust and reliable.

Data APIs

Data APIs can take on several forms.

For example, many third-party software providers, or Saas, give users access to their data. Often, at a cost. For example, Salesforce charges a $25 add-on fee per user, just to get access to their API (That’s if you don’t just pay for the enterprise version).

Your team can also decide to create an API to provide metrics like Datadog and Google Analytics.

Having standardized metrics like this is great because users don’t need to spend time defining their metrics. Instead, your team can create reliable metrics that are easy to pull.

Reporting

When I refer to reports, I mean a general analysis that has some form of repeatable component.

Often as a consultant, we will develop reports that we provide on a quarterly or yearly basis for companies that aren’t in the form of dashboards.

Instead, these will be reports that require some deeper analysis and distillation along with some of the standard charts.

Developing Data Products

Data products take on all different shapes and forms. However, they all have similar goals.

These goals revolve around providing insights, better services and to help companies improve their operations. Making a great data product means having a clear set of questions you want your end-users to be able to answer with your product or clear goals that help provide information and better services.

This was the first of several articles we will be discussing developing data products in.

This is part of our current push to develop a guide to help companies of all sizes improve their data strategy.

If you would like to have the guide when we complete it, then sign up below.

Also, If you want to read more about data science, big data and analytics, then check out the articles below.

How Do I Modernize My Data Analytics Strategy Part 1

How To Prepare For A Data Engineering Interview

What Are The Benefits Of Cloud Data Warehousing And Why You Should Migrate

5 Data Analytics Challenges Companies Face in 2021 With Solutions

How Your Team Can Take Advantage Of Your Data Without Hiring A Full-Time Engineer