The buyer is the hard part
Conrad is a European B2B electronics retailer selling to professional buyers: procurement teams, engineers, and the people who specify and maintain factory automation. I own research and design for its digital platform, which runs €400M GMV with 1.8 million monthly active users in Germany across seven product teams.
That buyer behaves nothing like a consumer. Purchases get approved by someone else, the catalog runs to industrial scale, and the person using the product is rarely the person paying for it.
Recruiting them is its own problem, and it is the problem that sets the pace of everything else. The people who can answer the question are the hardest to get an hour from, because an hour of a maintenance engineer’s time has a line manager attached to it.
Five people, which is the whole argument
I inherited two people and hired the other three, across two European locations.
Five is enough to build infrastructure and far too few to take requests. That constraint is the design brief: everything the team makes has to be usable by people outside it, because there is no version of this where five specialists serve seven product teams by doing the work themselves.
The panel
From 200 people to more than 7,000 in two years, B2B and B2C.
That is the difference between a study being a six-week recruitment problem and a two-day one, and it changes what teams are willing to ask. A question worth six weeks has to be a big question, which means the small ones go untested. A question worth two days can be anything.
Consented data handling under GDPR, settled once with legal and valid for every study after. Negotiating consent per study is how a panel quietly stops being used.
The repository
Company-wide, in Dovetail, open to every employee, with more than 300 interviews processed through it.
A product manager in one domain can use what someone learned in another, which is the only way a research function smaller than the company it serves ever compounds. Without it every study is a presentation that happens once.
The people who use it
The panel and the repository are only worth their cost if people outside the team run research themselves, to a certain degree, so product managers and product designers are trained and mentored to do it.
Five specialists cannot interview for seven product teams. Five specialists who have taught seven product teams to interview can, and that is a different job from the one most research functions are hired to do.
Decisions
Settle consent once, at panel level. Planned: handle GDPR per study, as each legal review came up, which is the path of least resistance because no single study is worth a policy argument. Argued: a review attached to every study is a toll on every question, and the first questions it prices out are the small ones. That is the quiet damage. A team stops asking anything that is not worth six weeks, and most of discovery is not. Happened: one basis agreed with legal and valid for everything after it, which is the reason the panel gets used at all and does not just get admired in a slide.
Continuous feedback alongside the studies
CSAT, NPS and goal completion rate, so teams see what shipping actually changed without waiting for the next round of interviews.
A Voice of Customer program collects what customers say at scale and puts it where planning happens. I have run one before, including the licensing side, so I know the failure mode: a dashboard nobody reads. The work is making the signal land in a planning conversation, which means getting it into the room where the roadmap is argued and not into a report circulated afterwards.