Build or buy analytics? There’s an 8x difference

Tl;dr By purchasing a modern data solution like Mixpanel, companies save time and money, and are able to focus their internal resources on what they do best: building a better product. Maintenance is a major cost. If you’re an enterprise company with a mature product, you could be spending $740K on maintenance alone.

Building a product that people will love is hard. Today’s customers are fickle, and the market changes fast. The only way a company can stay ahead of its competition is to better understand its customers, how they behave, and what they need. Leading companies like Amazon, Intuit, and Uber make great products by understanding how their customers use those products. But to inform decisions in all departments, data needs to be accessible to anyone seeking an answer.

We’ve spoken with a lot of companies that are struggling to find value in traditional analytics solutions. These companies need solutions to be more accessible, cost-effective, fast, and customizable to better report and act on the metrics that they care about. And here they encounter a dilemma they’ve faced in one fashion or another for years: build an in-house solution versus buy a software platform.

Building and maintaining a tool to handle hundreds of millions to billions of user behavior events isn’t a task to be taken lightly. We’ve devoted an entire report to the costs incurred in the internal tool dilemma, from building to running to using, but one of the major money pits is pure overhead.

The maintenance of such a tool will depend on how many user actions you’re logging annually. And no matter how many you log, this cost won’t be an afterthought. Let’s walk through it.

There are web services needed for ingesting and storing each user action, as well as the computational cost for each time the company queries the large set of data.

On the bright side, today’s web services are better and more affordable than ever before. On-demand pricing for a data warehouse offers annual effective prices (depending on precise ingestion and warehousing needs) that can bring the cost of each user action down to a fraction of a penny. However, a product with millions of users can easily scale the cost of ingesting, storing, and querying terabytes of user actions to more than half a million dollars a year.

And while the cost of warehousing unchanged data drops over time, as terabyte after terabyte of data accumulates, storing all of the data becomes more and more expensive. This can result in companies throwing away historical data rather than paying for storage. And this can result in a company not having the necessary data needed to make an important product decision.

Retaining the data necessary to make important product decisions, the cost of storage for an enterprise company could be as much as half a million dollars. And that’s just for ingestion, warehousing and querying costs.

In addition, there is the engineering cost of maintaining the internal tool. The build team will not be able to just create the tool and walk away. Employees will run into edge cases where the tool does not perform as expected. It is very reasonable to assume that the build team will spend a minimum of three to five weeks a year maintaining and updating the internal tool.

Taking into account market-competitive salaries for data scientists and engineers and the time they would devote in a year, an enterprise team might spend a quarter million dollars having its talent maintain an internal tool instead of focusing their efforts elsewhere.

Then, there’s time. Speed of iteration is a crucial aspect of building a better product. In a quickly changing market, companies need to move fast. With an internal analytics tool, implementing new product tracking doesn’t just cost money, it costs time. This slows down release cycles and hinders product improvements.

Of course, maintaining the tool is just one cost of doing business. For an internal tool, you need to build it first. Then, once you have it up and running, there are additional steps to make it usable.

If you want to get the maximum value out of your internal tool, you’ll need to equip it with a visualization layer so that decision-makers company-wide can put the data to use.

Ultimately, the cost of building an internal analytics tool is threefold:

• An initial cost outlay
• Maintenance costs
• The cost for a visualization tool on top of the infrastructure

Every company should make its decision according to its own budget and what’s necessary for its particular product. But no one should be forced to do so without having the costs laid out for them.

Our new report, Product Analytics Solutions for the Enterprise: Should I Build or Buy?, will take you through the dollars and (ten thousandths of) cents of choosing an analytics solution. Detailing the price of an internal tool from conception to daily use, this report will help product people assess the right analytics solution based on four factors: accessibility, cost-effectiveness, speed, and customization. With this whitepaper, you’ll be able to weigh each of these factors, invest in a solution accordingly, and get to building your product.