Planning customer discovery

Depending on the project’s level of uncertainty and complexity, the customer discovery phase can take up to several months and may need to be extended in case of any controversial findings. Every case is unique, but for an early-stage startup a 2-month discovery plan could look like this:

1. Start with desk research, personas modeling, and assumptions
2. Talk to advisors and subject matter experts
3. Recruit participants and schedule interviews 1–3 weeks ahead
4. Prepare the interactive prototype
5. Prepare the research goals and interview guides
6. Conduct interviews
7. Transcribe interviews, extract insights, look for themes
8. Prioritise findings
9. Make improvements to the product vision, UX, or feature list

Precise planning is the key if you want to avoid re-work and validate a lot of things before going into the development phase.

Stating your assumptions

Before starting the discovery phase, you need to list everything you already know and want to validate with customer interviews. You can use one of the idea validation frameworks, such as Opportunity Canvas, to list all your assumptions about the market, the business model, and your customers’ needs.

Personas modeling is also essential at this stage as you’ll be building your interview participant screener based on your personas and gaps in your knowledge about them. Start the discovery phase by conducting a Persona Workshop with your product team and stakeholders.

Press enter or click to view image in full size

For complex domains and B2B products, it is best to talk to the subject matter experts first— they’ll fill you in about the needs of your market and the taxonomy of the industry. This will save you a lot of time and costs, otherwise spent on figuring it all out during the interviews, or even worse — interviewing the wrong types of customers.

Besides, you’ll be prepared to speak the customer language — this will help you build rapport quicker and uncover deep motivations, rather than barely scratching the surface of a vast topic.

In longer terms, the same effect can be achieved via doing very thorough desk research on the subject — use a database like ResearchGate to learn what’s already known in your field.

Recruiting participants

If you don’t have your own research participants, you might need to use one of the user recruitment services available on the market, such as UserInterviews, Respondent, or Ethnio. For hard-to-reach audiences, it’s best to request participants from a specialised user research panel. I can recommend Borderless Access, as they ensure the quality of participants and take care of the screening process and interview scheduling.

When recruiting through a third-party agency, you’ll need to provide them with the screening criteria, which you can build with a simple Google form. The screener will also help you to obtain some basic information about your participants beforehand, so you can prepare better before the interview.

Press enter or click to view image in full size

Preparing for interviews

Once you’ve recruited the participants and scheduled the interviews, you can start preparing interview guides and research goals.

Research goals

Your research goals are the assumptions you’d like to validate and the most important things you’d like to learn from the participants. If you have multiple personas, prepare research goals for each of them. For example, you may formulate goals like these:

• “To understand customer’s motivation to use…”
• “To learn how customers do…”
• “To get acquainted with the language customers use to describe…”

Interview guide

The topic of preparation for an interview deserves a dedicated article (I really recommend Rob Fitzpatrick’s “The Mom Test”, an awesome book about customer interviews), but here I’d like to mention the main principles:

• Prepare open-ended questions so you’ll be doing little talking and a lot of listening.
• Avoid priming questions, such as “What is more important for you, scheduling your day or answering emails?”.
• Avoid hypotheticals that stimulate the customer to use their imagination rather than recall the facts.

The script also depends on the interview format you’ve decided to follow. The most common types are structured, semi-structured, and unstructured interviews. Choose the format based on how much you already know about the subject in general:

Press enter or click to view image in full size

If you’re only beginning to learn what’s important for your customers, the best way to reveal a lot of insights is to use an unstructured interview, which has no predetermined script to it. However, this type is the most difficult one to conduct, as you’ll have to come up with non-priming questions on the go.

That’s why most researchers recommend a semi-structured format for the discovery phase. You‘ll only need to prepare an “interview guide” rather than a rigid script for this one. The semi-structured interview allows you to stay within your research goals boundaries and improvise with a lot of probing questions to uncover deep customer insights. It can also be combined with a user testing of an early prototype, so you can build your discussion around the product interface.

During the interview

Technical aspects

It’s extremely helpful to record the interviews for further analysis, and this is something you need to notify the recruiting agency of. As most interviews are now conducted remotely, we all have the benefit of recording the sessions with video conferencing software. This removes the risk of losing important details and the necessity of inviting a note-taker to each live session.

For the sessions to run smoothly, take care of the technical details in advance:

• Tell the users not to call in from their mobile phones if a prototype is shared.
• Get prepared for the fact that participants wouldn’t know how to share their screen. Make a few screenshots for them to explain the conferencing software they’ll be using.
• When combining a user interview with testing of an interactive prototype, prepare to explain the prototype’s limitations to the participants and to your team. Prepare a back-up scenario if a user can’t open the prototype or the network connection wouldn’t allow them to.

Warm-up, then get more specific

At the beginning of the interview, allow time for the participant to get focused by starting with more general questions. Leave specific and personal questions for the second part of the interview, when you’ve built rapport and dived deep into the topic.

Guide the discussion

A participant may start digressing away from the topic in a semi-structured interview. Sometimes this digression is just small talk and isn’t related to your research goals — keep it up for a minute to build rapport, then softly pivot back on the topic you need to discuss.

However, if the digressing topic keeps re-emerging, it’s a clue to a deep customer's motivation or a need that you didn’t think to ask about. Pay attention if the participant’s getting emotional when they’re talking about it.

Transcribing the data

When it comes to transcribing, I can recommend two products: Otter and Descript. Descript’s pricing policy is a bit more flexible, so I’ve decided to stick to it. Both products are AI-powered transcription services, which means the quality is not quite as accurate as a human-created transcript. However, if you’re transcribing the interviews you’ve conducted yourself, you won’t have a hard time interpreting the inaccurate parts.

Still, if you need a human-created transcript, there are several options in the market now, such as Rev and Reduct.

Analysis of the interviews

When you have transcribed the interviews, the next step is to analyse the data you’ve collected. Even if you’re pretty sure you’ve already grasped the essence of the interviews, it’s worth doing a proper analysis, for two main reasons:

1. The interview situation is very intense, and it’s easy to overlook things that don’t fit with your preconceived assumptions.
2. Your team and stakeholders will understand how you reached various conclusions, and your product decisions will be justified.

Extract insights

Start the analysis by reading through the transcripts and colour-coding fragments that are related to your research goals. The aim here is to reduce the amount of raw data and highlight only the most important insights. If you’re at an exploratory stage and don’t have a clear idea of what patterns you are searching for, pay attention to what was important for the participant. It could be something they were particularly enthusiastic about. To remind yourself of these nuances, it’s helpful to re-watch some of the interviews and observe the emotions.

After you’ve colour-coded the transcripts, extract the highlights into an insights database. I use Notion for insights management as it allows creating relations between data fragments and enables grouping of insights by participant or priority.

🔗 Get Notion template for the insights database here