Categorizing design decisions in enterprise product development
It’s an interesting time to be a data practitioner. Never before has the market followed the needs of data teams so closely and intently. At the same time, the growth and evolution of the data infrastructure and tooling market has created grounds for critique, hot takes, and calls for action.
As I go about planning for the short-term and long-term functionalities of General Folders, I find it helpful to identify the design decisions that attract and keep the attention of customers. By categorizing these design vectors, my hope is to bring structure to the exploration of the infinite space of product variations to discover and build products that sell.
The decision vectors that I was able to uncover are configurability, functional variety, application universality, and audience universality. Together they help shape the user experience of enterprise products. It is important to note that these vectors are not orthogonal, meaning that moving on the spectrum across a single vector, can potentially impact where one stands within another. Without further commentary, I present four design vectors and discuss some of the implications of moving across each vector.
1 Definition
Configurability is used to define the degree to which one company’s version of a product or service can differ from another company’s version. Configurability is achieved by exposing programmatic or low code toggles that allow enterprise customers to modify the product or service.
2 Examples and best practices
Box, Dropbox, and GitHub offer very few toggles, whereas Workday, Salesforce, and SAP offer many.
Although larger customers require heavily customized solutions, customization comes at the cost of simplicity in workflows and onboarding. Salesforce has recently launched Salesforce Easy to simplify and speed up the onboarding process for their customers.
A Dell or HP laptop is more configurable than a MacBook. Yet many companies still choose MacBooks for their employee base. They choose convenience over configurability when configuration does not provide clear benefits for their use cases. It’s nice when you can skip worrying about antivirus software.
1 Definition
Functional variety refers to the span of distinct functionalities that are bundled into a sellable unit, wherein each functionality can realistically be removed, shipped, and sold as a standalone product or service.
The functionalities included and those excluded in the final sellable unit define its bundle boundaries. This choice can go far beyond the product and become the boundaries to which the company adheres.
2 Examples and best practices
A product can be built such that it serves a single function very well. Or it can be a Swiss Army knife amalgamation of many functionalities. Consider for example, Databricks. It’s a bundled offering spanning data warehousing, data streaming, and ML Ops functionalities. Another bundled service, Notion, includes note-taking, ticketing, and data management functionalities inside a single sellable unit. At the other end of the spectrum, there is a lightly bundled offering like Fivetran, which provides SaaS data integration services.
Note that bundling stands in contrast with multiple business units delivering services that are sold separately. Atlassian, for example, allows customers to pick and choose. Paying for JIRA does not get you a Confluence membership. Similarly with AWS, paying for Redshift does not get you access to SageMaker. On the other hand, Notion and Databricks bundle all functionalities into a single sellable unit. The same is true for Monday.com and ClickUp, which are considered to be under the Work OS umbrella of products.
In a brand new space, the market maker has the benefit of defining the bundle boundaries. They are able to shape customer expectations. For example, JIRA has come to define what it means to be a ticketing system. GitHub has done the very same for a version control service. A new company has no choice but to implement core functionality as set forth by the market maker. In order for the next JIRA to succeed, GitHub integration is a necessity. The next Workday has to support seamless management of international payroll.
In some markets, the agreed upon bundle boundaries are under development. BI tools are such a product — the market has yet to figure out the ideal set of functionalities that produce a winner.
Opportunities for bundling occur when an incumbent makes it difficult to buy or to install and integrate their product. But this is not a sufficient condition to bundle. When bundling doesn’t fully replace the need for an existing product, the company’s customers would still need to buy that product — regardless of all the hurdles it may introduce.
When the sufficient conditions for bundling are not met, bundling is not a good use of a company’s budget. Instead, it is smart to provide integrations with existing products and services. A look into Snowflake’s list of technology partners, displays the functionalities they choose not to build but for which to offer seamless integration.
Ultimately, it’s easiest to sell a product when the bundle boundaries are shaped by the market. A product makes sense in the context of adjacent products. It lives in conjunction with all the existing workflows that it does not affect. Bundle boundaries are inevitably shaped by the ecosystem in which the product is to exist. Study the workflows, then fill in the gaps. A good example for this is Square which aims to address all the needs of enterprises that operate as a click and mortar business.
The final note here is about managing bundled products. Attention to the proper allocation of people to each functionality is key. As a company begins to neglect any one of its bundled functionalities, it can lose customers as better unbundled services consequently emerge. A good example is ticketing in Notion, which is being unbundled and built anew by companies like Linear.
1 Definition
Application universality is a term to describe the degree to which a single product can be applied to varied use cases. A product with high application universality can be considered a horizontal tool or a generic tool.
2 Examples and best practices
A great example of a universal product is Excel. And Gmail. You can use Excel for project management, for financial modeling, and for accounting. Gmail can be used for mailing lists and newsletters, for drip marketing campaigns, and for cross-company and internal communications.
Application universality is commonly confused with functional variety. For example, Excel is consistently considered a bundle (see “the unbundling of Excel”). However, it is more accurate to consider Excel a single product with high application universality rather than a bundle. A bundle is when you can break and remove parts that can be sold separately. Excel can be used as a CRM and for accounting. But you can’t remove the CRM functionalities of Excel while leaving the accounting functionalities intact. Consider, also, the kitchen knife. Unlike a Swiss Army knife, the kitchen knife is not a bundle. It has application universality but lacks functional variety.
Horizontal products, i.e., products that don’t target any one particular industry, are intended to have high application universality. For example, AWS S3 is used to store a company’s data—regardless of its industry. A product with high application universality is naturally able to target a larger market. Having said that, vertical focus doesn’t necessarily reduce a company’s ability to grow. Vertical focus is an opportunity to target a bigger part of a smaller but neglected market. As an example, consider Shopify which builds for online retailers.
It is important to note that application universality is not always ideal. In the case of Powerpoint, for example, it can lead to confusing user experiences. It’s not clear whether Powerpoint is a way to deliver a report intended to be consumed asynchronously, or a medium to deliver a speech in real time. Similarly, it’s not clear whether Slack is meant to fulfill the same role as email (asynchronous and high-context workplace communications) or a messaging app (synchronous and low-context workplace communications). A final example is in the BI tooling space. It isn’t clear whether a BI tool is intended for data analysis or if it is a medium for delivering reports to the rest of the company. Are dashboards an intermediate artifact meant to be used by data practitioners? Or are they meant to be consumed and explored by business users? This lack of consensus leads to confusion and potential misuse, for example, when business users take action in response to noisy data.
1 Definition
Audience universality is used to refer to the the variety of user personas that are able to employ and benefit from a product in the enterprise.
2 Examples and best practices
Google Docs and Gmail have high audience universality whereas AWS has low audience universality. Every employee gets a Gmail account whereas only a subset of engineering team members get an AWS account.
Supporting diverse user personas doesn’t come naturally to all product categories. Enterprise roles vary greatly in their responsibilities, required skills, and work processes. This in turn impacts the ideal set of features for a successful product. For example, BI tools that optimize user experiences for business users tend to underinvest in the features needed by the employees who spend the most time with the tool: the data practitioners. In this particular scenario, and given that data teams are the buyers of BI tools, neglecting their workflows and processes is a not the best sales strategy.
Yet another example where opening a product to more user personas is not straightforward is in enterprise messaging. Consider Slack, for example. Given that Slack invites the entire employee base to contribute to an always-on stream of conversations, it can become a challenge to manage workflows for employees who are on different schedules.
Infrastructure tools take pride in having low audience universality. Consider Terraform, for example. Most employees rarely know if their company is a Hashicorp customer. The same is true with Fivetran or Confluent. These products are designed in such a way so as to minimize the number of employees required to operate the product.
Where a product or service stands with respect to configurability, functional variety, application universality, and audience universality make it confusing or extremely useful. Making deliberate decisions across these vectors can lead to easier-to-sell and easier-to-use products, while ensuring the efficient use of capital.
Happy building,
Pardis
P.S. Special thanks to Ryan O’Neill, Scott Beru, Abhiram Ramesh, Shane Johnston, Preeya Phadnis, Pedram Navid, and Hassan Jaferi for reviewing this post and providing valuable feedback. Thanks also to those whose discussions and work has provided inspiration for this post.
