Most teams measure sprints by output. Stories closed, points burned, features shipped. The scoreboard rewards volume. But volume is a trailing indicator — it tells you how busy you were, not whether you built the right thing.
There is a different discipline worth trying: constrain the sprint to a single feature, then put it in front of a real customer before the next sprint begins. Not a demo. Not an internal review. A live session where someone outside your building tries to use what you built, and you watch.
Teams that adopt this pattern consistently make sharper prioritization decisions in the sprints that follow. The learning compounds faster than the velocity loss.
When a sprint carries three features, each one gets a third of your attention during review. Feedback blurs. You cannot tell whether the customer hesitated because of Feature A's wording or Feature B's placement. Causality dissolves.
A single feature gives you a clean signal. When the customer pauses, you know exactly where they paused. When they smile, you know what caused it. When they ignore the thing entirely and ask about something else, that silence is information — and it is unambiguous.
One feature also forces a harder conversation at sprint planning: which feature matters most right now? If your team cannot agree on the single most important thing to build this week, that disagreement is data you need to surface before you write any code.
Three questions help narrow the field:
Which problem has the shortest path to observable customer behavior? You want something you can watch a person use, not something buried three screens deep behind a settings page. Prioritize the feature that changes what a customer does, not just what the system does.
Where is your confidence lowest? If the team already knows a feature will work — maybe it is a direct customer request with obvious implementation — it does not need the feedback loop as urgently. Pick the feature where you are guessing. That is where learning has the highest return.
What would be most expensive to get wrong? Some features, once shipped, are hard to walk back. They set expectations. They change pricing conversations. They create migration paths. If a feature locks you in, it deserves the slower, more careful treatment.
You will not always find one feature that satisfies all three. Two out of three is enough to justify the pick.
The session is not a usability test with a script and a one-way mirror. It is simpler. You sit with a customer — on a call, in their office, wherever they are comfortable — and ask them to accomplish a task using the thing you just built. Then you shut up.
A few ground rules:
Do not explain the feature before they try it. The moment you narrate, you pollute the signal. If the feature needs a walkthrough to be usable, that is your first finding.
Bring someone who built it. Engineers who watch a customer struggle with their own work develop a different relationship with feedback. It stops being abstract. It stops being a ticket.
Write down the three moments that surprised you. Not the moments that confirmed your hypothesis. The surprises. Confirmation feels good but teaches nothing. Surprise is where the learning lives.
Do this before the next sprint starts. Not next month. Not in a quarterly review. Before the next planning session, so the learning directly shapes what you build next.
It will. The backlog always screams. Stakeholders will point to the roadmap. Sales will forward customer emails. Someone will mention a competitor.
Here is the reframe: you are not building less. You are building one thing well, then deciding what to build next with better information. The team that ships three features without feedback is guessing three times. The team that ships one feature with feedback is guessing once, then adjusting.
Over a quarter, the one-feature team often ships fewer total features but delivers more value, because fewer of those features need to be reworked, reverted, or replaced. The waste rate drops.
There is a real trade-off: some weeks, you will feel slow. Some stakeholders will lose patience. You have to be honest about that cost and decide it is worth paying. It is not a free optimization. It is a bet that learning compounds, and the evidence supports that bet.
Sprint one, you ship a feature and learn something about how customers actually work. Sprint two, that learning shapes what you pick and how you build it. By sprint five, your instincts are calibrated to real behavior, not internal assumptions.
Each feedback session makes the next prioritization decision slightly better. Slightly better decisions, repeated over months, produce a product that fits — not because you planned it perfectly, but because you corrected course early and often.
Velocity without direction is just motion. One feature, one session, one correction — repeated — gets you somewhere worth arriving.
Be the first to comment.
0 comments
Loading comments...