“The power and influence of leading customers is a major reason why companies’ product development trajectories overshoot the demands of mainstream markets.” — Christensen, C. M. The Innovator’s Dilemma
Nearly as important as listening to your customers is knowing when not to listen. Identifying customer feedback that, while loud and assertive, is not representative of your target user will focus your product and save thousands of hours.
Shutting out noise is especially important for startups on the cusp of explosive growth. It seems obvious to say that minority use cases shouldn’t fill most of your time, but the reality is harder to navigate. Switching gear from surviving as a startup to thriving as a business requires a dramatic rethink of your relationship with individual customers.
Startups are built on strong bonds with early adopters. In the first few months of existence, you’re grateful to have a handful of users. They are a precious resource — the special few who have enough faith in your vision to give you their custom, test drive your features and recommend you to friends.
Brian Chesky would routinely appear on the doorsteps of Airbnb’s first hosts in search of feedback. Flickr’s first users were personally greeted by co-founder Caterina Fake (who went on to co-found Etsy and Kickstarter, among others).
With ingrained dependence on early customer feedback, and such pedigree to support that approach, it’s hard to kick the habit and start closing your ears to some of the voices in a growing community.
Lending Works, the peer-to-peer lender I work for, is faced with exactly this challenge. Its lending platform, through which investors lend their capital directly to borrowers, is scaling to support a loan book growing 100% every year. Its user base, however, is still one of early adopters who are accustomed to having their voices heard and personally acknowledged.
Lending Works allows every lender to export their full loan book, containing details of the capital and interest spread across every repayment of every chunk of every loan to which they contribute (for diversification’s sake, no lender ever funds a loan in its entirety). It was a feature developed in the name of total transparency. The full data, as seen internally, is available to anyone who wants it.
Recently, however, Customer Service received an email from a disgruntled lender who had been running her own interest calculations. After pulling her Lending Works data into a personal spreadsheet and crunching the numbers with macros she’d written for just that purpose, she came to an alarming conclusion: we had underpaid her.
Specifically, she thought we’d underpaid her by £0.00005. Five thousandths of a penny.
Daft? If only. Our algorithms have been tested every conceivable way — automatically, by hand, upside down, underwater. We knew our calculations were correct but, in finance especially, customer trust is so essential that we had to prove they were correct. It was serious enough for our Director to tackle it personally.
On investigation, he discovered that the borrower to whom one of this lender’s chunks was assigned had, unknown to the lender, changed their monthly repayment date. Without this information, manual calculations would be thrown off.
We presented this discovery to the lender. In response, we received no fewer than 9 equations the lender hypothesised we were using to calculate interest. None produced the value she expected, even accounting for the date change. Our valiant Director got his head down in the decimals once more.
Hours later, a simple explanation was found. Because the borrower’s nice, round loan amount was split over more than 800 chunks, each belonging to a different lender, some rounding was required even at the 5 decimal place level. On some chunks you would end up a few thousands of a penny down, on others you’d be rich to a similar tune. No one ever loses out. The system was working as intended.
We created this customer’s problem. By giving lenders access to 5 decimal places worth of data, we had hoped for transparency. Instead, we gave power users a partial, unsatisfactory look into internal processes that left them with questions. The only solution to those questions was to give them yet more information.
Rounding the available data to whole pence would prevent this from happening. Pounds and pence are all the vast majority of our user base care about. But changing the status quo risks angering those edge users who excel at making their voices heard.
Individuals who consume disproportionate amounts of executive time despite their niche use cases are not unusual. My all-time favourite customer exchange was with one honorable gentleman uncontrolled in his indignation that nowhere in our sign-up form’s title dropdown did we list his rank within the British peerage. For this man, ‘Mr’ wasn’t merely a demotion, it was an affront to Her Majesty the Queen.
Did we change our title dropdown to keep a customer? We did not. And, faced with any similar situation, neither should you. As your startup’s growth engine starts to turn in earnest, direct your effort towards the health of your community, not lone, unrepresentative customers.
Identifying who constitutes your community — who to nurture, and who to let fall away — is the first step towards healthy growth that won’t leave your small team struggling to serve the eclectic whims of a noisy minority.
Whatever user research you’ve done previously, I’m sure you’ll have intuitions about what constitutes ‘normal’ behaviour. For Lending Works’ investors, ‘normal’ might be a few thousand pounds, lent out automatically at the highest interest rate and longest term, with only occasional logins to check on investment growth.
Validate that intuition. Conduct in situ interviews with a generous sample of your user base, or, if that’s beyond your means, survey an even larger sample. Be careful not to lead your participants. Ask only about their past behaviours, never their future intentions.
Next, size up the portion of your user base who fall outside that set. If a large minority (more than 20%) share a uniformly ‘abnormal’ behaviour, congratulations: that’s true product discovery. Unfortunately, you must now consider how to unify the needs of both those groups. They’re here to stay.
In the more likely event that the minority comprises many fragmented, outlying behaviours, or the uniform minority is small, it’s time to phase them out. It will go against every impulse you possess, but it must be done for you company’s health and your personal sanity.
Doing it is delicate work. Do not rip the plaster off. Do not start disabling features and making grand announcements. RateSetter did this recently, switching off the rate-setting feature for which it’s named on the grounds that just 7% of investors used it. Unfortunately for RateSetter, that 7% were loud and unified in their condemnation of the change. The feature was swiftly re-enabled.
In these situations, gentle deprecation is the only viable approach. Switch off the feature for new users. People are loss-averse, but they don’t miss what they never had.
In the next phase, warn any existing user who voluntarily turns off the feature that they won’t be able to turn it back on. You will, I guarantee, receive a trickle of complaints, but no coordinated front seeking justice.
When the proportion of active users still clinging to the old ways becomes small to the point that any semblance of support is too much to justify, then, and only then, is it time to pull the plug.
As your company prepares to explode, beware the minority needs that start taking most of your time. Grow a healthy community around defined use cases and feature sets, and accept when you need to let the rest down gently.