PS1: Sourcegraph is hard to learn and therefore hard to use

[benvenker]

Problem statements

We’ve observed that for new users, Sourcegraph's core search workflow is unintuitive and unfamiliar, which results in a steep learning curve and poor first experiences that are unlikely to result in a personal "aha" moment.

Similarly, we've observed that existing users often lack—and struggle to build—confidence in searching with Sourcegraph, which limits the value they get out of search. For these users where search is their primary entry into Sourcegraph, a poor search experience will make it less likely they'll explore beyond search and discover the full value Sourcegraph provides.

Developers being onboarded to Sourcegraph typically don’t personally have a big code problem (but their employer might), don’t understand how Sourcegraph can help them, and don’t want to waste time learning a new tool (they already have too many tools).

Some quotes from our customers can be found here.

Top problems

• Sourcegraph doesn’t behave as users expect based on previous experiences with search tools, which leads early frustration and low motivation to learn Sourcegraph, and may also reinforce preference for their existing or previous search tools. Specific misalignments with user expectations include:
  • Add examples
• Today, developers trialing Sourcegraph for the first time often need to make up a “cool” search to see what it can do because we don’t show them any value by default (we provide an open-ended capability not a tailored solution).

Additional context and constraints

• ...

Design challenge statements

• How might we...

[rrhyne]

Some quotes from our customers can be found here.

This starts to be convincing, but it is a very small sample of feedback. I think it could be stronger and include quantitative data to avoid recency bias and negativity bias.

[rrhyne]

I love this issue!! It covers so many needs, and easy to consume.

The one thing I think is missing is a design challenge on showing value. Ben mentions it with 'the aha moment', which is to this point. But how are we going to make sure we address that the user must see enough value up front to compete against the IDE as a search tool, or invest time in learning even an improved experience?

How would we measure that we've done that effectively?

[emchap]

This work is probably separate, but just want to add to the issue if useful—our 101 training for users is stored at https://github.com/sourcegraph/customer-training/tree/main/trainings/sourcegraph-101 , if you want to see the things CE hits on to enable the customer early in their journey. (All other trainings available at https://github.com/sourcegraph/customer-training/tree/main/trainings .)

[quinnkeast]

On further reflection, this problem statement was doing two things in one: capturing a large, open-ended problem that will take some fundamental work to solve, and also framing a short-term gap with simple UI. Solving the latter won't solve the former, but the solution to the latter is informed by the same high-level problem statement.

I've broken out the short-term effort into #39817 and we'll refine the problem statement content here to better focus on the large, open-ended problem space.