Title Image

Blog

How much is my data product worth? By Sarvenaz Rahmati & Alexandre T’Klint

  |   RSS Feeds

How much is your data worth? The same question arises over and over again, “what is the value of our data products?” Our team can try to answer this question by tracking estimators such as usage and consumer metrics. But, the truth is – no one really knows what a data product is really worth. …
The post How much is my data product worth? appeared first on Collibra.

How much is your data worth?

The same question arises over and over again, “what is the value of our data products?” Our team can try to answer this question by tracking estimators such as usage and consumer metrics. But, the truth is – no one really knows what a data product is really worth. 

Mike Ferguson, Managing Director of Intelligent Business Strategies Limited, explains that the term ‘data product’ is used to describe everything that provides insights from data. Including anything from virtual data assets to SQL queries to reports and APIs – and everything in between.

Once you’ve factored in all the tangible measurements, you can’t ignore the abstract influences.  An essential criterion to valuing a data product is to determine if the data team’s resources were used effectively.

Yes, a lot goes into figuring out the value of your data products, and you probably won’t capture everything, but if you measure most costs and components, it’s well worth the effort. An effective data product leads to accurate decisions.

We recommend starting with a specific and popular data product within your organization.  In this blog, my team shares our experiences and the steps we took to determine the value of one of our data products.

Kickstarting the research 

We kickstarted the research by focusing on a specific data product called the Data Intelligence Usage Dashboard. A dashboard that helps our sales engineers understand how they can guide users through a Collibra proof-of-concept (POC). If used correctly, the sales engineers can improve a customer’s user experience and satisfaction during a POC.

To evaluate the dashboard, we focused on costs, revenues and net value. As a first step, we estimated the cost of all the resources that the data product consumes, then we measured the amount of money that this data product is returning to our company. And finally, by comparing costs and revenues, we evaluated how profitable or unprofitable this data product is. In image 1, you can see the general model for valuing the data product.

Image 1: General model for data product valuation

1. Costs

To evaluate the costs, we first determined what resources the data product uses. In image 2, you can see the costs of the data intelligence usage dashboard in its first six months for each cost parameter. 

Image 2: Costs of the data intelligence usage dashboard in its first six months

1.1. Amazon Web Services (AWS) cost

When creating the Data Intelligence Usage Dashboard, several AWS services are used. For example, the raw data is stored in S3 buckets, EC2 is used for cloud computing and Redshift is used for data warehousing. In image 3, you can see what Amazon Web Services are used to create this data product.

Image 3: Using AWS to create a data product

In addition to the Data Intelligence Usage Dashboard, many other data products in Collibra use the same architecture. Therefore, the challenging part was calculating how much AWS resources were used by this data product and thus what fraction of the costs can be related to it.

To simplify the problem, we allocated the costs based on runtime with the assumption that every unit of runtime consumes the same amount of resources. In image 4, we described the resource allocation assumption in this research.

Image 4: Resource allocation assumption in this research

According to the previous assumption, we calculate the cost of S3, Redshift and EC2 for the Data Intelligence Usage Dashboard.

1.2. Creation cost

The creation cost is the cost for all people involved in creating the data product. In this case, we needed the help of one data scientist for designing and implementing the data product, one data engineer to build, test and maintain the data pipeline architecture, one pre-sales engineer to ensure the data product is meeting their needs and two senior managers for giving feedback. 

1.3. Maintenance cost

Maintenance cost is the cost of modifying a data product after delivery to correct faults or improve performance. In the case of the Data Intelligence Usage Dashboard, the maintenance cost is the cost of people who are working to maintain this data product once it has been delivered. Since this data product fails very little, the maintenance cost is very minimal.

1.4. Tableau license cost

The company has to buy a Tableau license for all the people who create or use this dashboard. So, we also considered the cost of these licenses.

2. Revenue

In the second step, we intended to find the revenue for the proof of concepts in which sales engineers used the Data Intelligence Usage Dashboard. This wasn’t easy to define. Do we look at the number of views on the data product? The number of closed deals per user of the product? We decided to go for the interview approach and ask the sales engineer how much they value the dashboard. 

When we interviewed the sales engineers, they all agreed: “This data product helps us but it’s a tool that only supports the process. It only contributes to part of the value created.” Therefore they estimated the attribution rate to, on average, be 4%. Meaning that 4% of the revenue that was generated in the past 6 months could be attributed to the existence of this data product. 

3. Net Value

After calculating the costs and revenues based on the assumptions made in the previous sections, we subtract costs from the revenue and evaluate the data product net value. 

If the result of this subtraction is positive, the data product is beneficial for the company and we can keep it the same way or improve it further. If the result is negative or equal to zero, it means that the costs of this data product is similar or more than the revenue, and the data product is worth nothing. So, we have to search for the weaknesses. Maybe the data product needs improvement, or there’s a need for enablement within the company. Maybe no one knows about the data product. 

Conclusion

A data product can help you make the right decisions. An unprofitable data product wastes the costs and resources of the organization. It should be modified so that it becomes profitable. If this is not possible, it is better to sunset the data product. Hence the importance of the monitoring step in the 8 step process to create a data product; the monitoring phase values if the data product can be kept as is, should be enhanced or decommissioned. 

Based on the analysis done in the previous sections, there are two possible ways to increase the net value of a data product: reducing the costs and/or increasing the revenues. 

Reduce costs 

Since, in this case, the most significant chunk of the costs is related to the creation, we have to reduce the creation cost. For example, have less people involved.
The compute cost on AWS can be decreased by optimizing the computation and requiring a smaller and cheaper instance or using spot instances as they are compatible with the nature of the job.

Increase revenue

Also, to raise the revenue of the Data Intelligence Usage Dashboard, the most effective way is to increase the attribution rate.
Have more people use it or know about it. 
Better enable consumers to use the data product.

At Collibra’s Data Office, we are constantly working on Data Intelligence, which means connecting the right people to the right data, insights, and outcomes. With this experiment, we’ve walked you through the steps to estimate the value of a data product via outcomes and costs related to maintaining and building it. Once you know how much your data products are worth, you can estimate the value of your data.

The post How much is my data product worth? appeared first on Collibra.