Feasibility spikes

Surfacing a hard constraint after design is locked and the launch is marketed is how trust dies. As soon as you smell trouble — data model mismatch, rate limits, legal hold, mobile offline — say it with a proposed check.

Time-box the spike. Bring back a written note: what you tried, recommendation, residual risk. That is the same muscle as a mini design doc.

Walking through it

I am 60% sure FraudSvc cannot support this webhook shape. Two-day spike starting tomorrow; if blocked we need plan B by Thursday before design finalizes copy.

The written note is the deliverable, not a prototype you accidentally ship. Name the residual risk so the trio can choose: proceed, cut, or wait.

Volunteer for discovery spikes in 1:1s. "What problem are we trying to move this month, and how will we know?" is the question that surfaces them.

When process is broken

Not every company runs empowered teams. You may get fixed roadmaps and tickets that pretend to be requirements. Still useful moves:

• Ask for the user problem in the ticket template
• Attach a one-paragraph "eng proposal" with a cheaper alternative
• Instrument even if nobody asked — so the next debate has data
• Build credibility with reliable delivery; spend it on shaping the next bet

Do not burn out fighting the org chart every week. Pick one collaboration improvement per quarter. Exit or transfer if you are permanently blocked from learning product sense — that skill compounds for SDE2 and later, and is hard to rebuild from pure ticket work.

Common mistakes

• Raising the blocker in sprint review. Too late. Say it when you smell it.
• An open-ended spike with no Thursday. Time-box, or it becomes a rewrite.
• A prototype that ships by accident. The note is the deliverable.
• Fighting the org chart every week. One improvement per quarter; then decide if you stay.

Check yourself

1. You are 60% sure a vendor webhook will not fit. What do you send today? A time-boxed spike and a date for plan B, before copy is finalized.

2. Why is the written note the deliverable? Stakeholders need a recommendation and residual risk, not a surprise prototype in production.

Key takeaways

• Raise feasibility risks early, with a proposed check and a date.
• Time-box. Come back with tried / recommend / residual risk.
• Pair a blocked result with a cheaper path.
• In a feature factory, still attach a one-paragraph alternative.

Further learning

Next, partner on the edges in working with designers.