Customers are idiots when they're telling you what they want. Instead, observe their behaviors and take in their holistic behavior. You can then map out their motivations, which is much more telling than what they tell you.

Why is this such a problem?

Customers only know what they have experienced. They lack the complete market context and cannot imagine what they simply do not know about emergent technologies and market trends. If you ask your customer for solutions, they will respond only with their current experience. This does not fulfill the vision that you have set out for yourself or your company.
Listening too closely to your customers puts you in a pitfall of making incremental rather than bold improvements. This results in "me-too" products and missed opportunities on not seeing the ultimate outcome for customers. At the time, all one sees are solutions that solved immediate, short-term pains. One should be asking "what is it that they really want?"
Listening too closely to lead users (sophisticated users) leads to complex solutions that deviate from normal market use.

What is the solution?

Plan outcome based customer interviews by carefully planning out the product usage process that includes the "successful" outcome of each step
User selection is highly important. Who you select will determine the life or death of your product. Select diverse users who can judge the value and [specific characteristic] of your product.
Hone in on outcomes and solutions to achieve those outcomes and weed out anecdotes, vague statements, and other irrelevant comments. Dig beneath the surface of the customer's words, ask clarifying questions, and ensure that the participants consider ever part of the product usage process. Redirect offshoots to ensure that the customer is thinking critically about the underlying process (what frustrations do you encounter? what would be the ideal situation?)
The outcome should be the "type of improvement" (increase / decrease) and a "unit of measure)" (time / number / frequency) so it can be measured.
Use opportunity matrices to figure out segmentations and solutions.

Some question pointers:

Suss out if the user cares about the outcome.
- Have they tried to solve the problem by paying or using [solution]?
- What are you trying to do?
Suss out their frustrations.
- What is difficult about [process]?
- What do you find frustrating about [process]?
- If you had a magic wand what would this [process] look like?