Startups are successful in the early days usually for one of two reasons. One is having a unique insight or pain point in the world that you want to solve (usually for yourself as the founder first), and assuming it is pain experienced by others. The other is to listen to customers, deliver value to those customers, and make sure they understand and appreciate the value you’re providing. The second way requires founders to hone specific skills in the early days of a startup — which can actually make it harder to scale out of the early stage, but pays off with sustainable growth in the long term.

The people that try your product early on see the potential of your product and are willing to forgive flaws — at least for a while. They have done an incredible amount of work to make the product work for them. By most definitions, they become “power users.” These power users are heavily engaged with your product, but they also deliver a ton of feedback on how the product could be better for themselves.

The early employees of your company tend to be very similar to your early customers. They (hopefully) use the product quite a bit, and joined the company because they understood the long term vision. These employees then start recommending and building products for themselves also, especially in consumer businesses. Everyone is excited to build these features because employees want them and existing customers want them, so the company builds them. The features get built, and there is no impact on growth of the business. Our partner Sarah Tavel talks about this in her lessons from scaling Pinterest.

Why is that a problem? Well, in an old Quora question someone asked, “What are some of the most important things you’ve learned in marketing?”, and my reply was “You are not your customer.” As a company employee, even if you look exactly like the early customer, and you built the product for people exactly like you, you have way too much domain knowledge to truly represent the long term customer. Your early users are also no longer the customer. Both employees and early users have have built up too much domain knowledge.

Your customer focus should always be on new or potential users, not early users. Early users will bias experiments, prompt you to build more and more niche features, and stunt growth. Power users can’t be much more engaged, so building more things for them doesn’t usually help the business. It does, however, make the product harder to understand for new customers. Sure, you have to do enough to keep these power users happy enough to stay, but the much more daunting and important task is to find new people to delight, or to figure out how to delight people who weren’t initially delighted by your product.

This post originally appeared on the Greylock blog.

Currently listening to Ambivert Tools Volume One by Lone.