Blog

Handling unplanned work without wrecking the sprint

How to measure unplanned work, budget for it in planning, route it through one person on rotation, and say no to mid-sprint requests without a fight.

Your team leaves sprint planning with a clear goal and a sensible amount of work. By Wednesday, a production bug has eaten a day, sales has asked for "one small export", two people have spent their mornings answering questions in chat, and the sprint board looks like a wish list. At the review, the team says it was interrupted all sprint; stakeholders hear that it missed its commitments again. Both are true, and nobody can say how much of the sprint went where.

Every team that runs software in production gets unplanned work. The problem is handling it informally: everyone picks things up as they come, nobody counts it, and the plan pretends it doesn't exist. Four habits fix that: measure it, budget for it, route it through one person, and decide quickly what waits.

Measure it for two sprints first

You can't budget for something you've only got a feeling about. Before you change how you plan, spend two sprints counting. Keep it light, or people won't do it:

  • Any piece of work that wasn't in the sprint plan and takes more than about 30 minutes gets a ticket or a card, tagged unplanned.
  • Each one gets a category and a rough time in hours, filled in when it's done, not estimated up front.
  • Shorter interruptions (a quick question, a two-minute fix) get a tally mark against the person's name in a shared note. You want to know how often people are pulled away, not just for how long.

Use three categories. They behave differently, so you'll handle them differently:

Category Examples Typical shape
Incidents Outages, pages, data problems, security fixes Urgent, unpredictable, can't wait
Requests Bugs reported by users, small asks from other teams, access and config changes Steady flow, mostly can wait a day or two
Questions "How does this endpoint work?", "Can you look at this log?" Frequent, short, expensive mainly because of context switching

At the end of two sprints, add it up. Say a team of five (hypothetical) logs 70 hours of unplanned work across two sprints, plus around 120 tally marks. That's 35 hours a sprint, and an interruption every hour or so for someone on the team: a number to plan with and a list to work on, instead of "we get interrupted a lot".

Budget for it in sprint planning

The PTMP study guide works through a capacity calculation that subtracts time off, meetings and operational work before applying a focus factor. Most teams skip the operational-work line or guess it. Use your measured number instead.

Continuing the example: five people, two-week sprint, 400 hours in total. After time off and meetings, say 340 hours remain. The log says unplanned work runs at about 35 hours a sprint, so take that out explicitly: 305 hours. Then apply the focus factor for context switching.

Two rules keep the budget honest:

  • Use the measured average, not the best sprint. If the last four sprints had 20, 35, 30 and 50 hours of unplanned work, plan for about 35, not 20. If the numbers swing wildly, plan for the higher end until you understand why.
  • Say the budget out loud. In planning, state it: "We're holding about 10 percent for incidents and requests." When a stakeholder later asks why the team didn't take on more, you can point to a number that came from data, not caution.

At the sprint review, "we got interrupted" becomes "we planned 35 hours for unplanned work and used 48, mostly on the billing incident", which people can act on.

Put one person on interrupts

If unplanned work goes to whoever notices it first, everyone's day gets broken up. The chapter "Dealing with Interrupts" in Google's Site Reliability Engineering book (2016, written by Dave O'Connor) makes the case for the alternative: give interrupt work to a designated person for a period, and let everyone else focus on project work. It argues that the same person shouldn't be expected to carry on-call duties and also make real progress on project work, and that people who aren't on interrupts shouldn't quietly work the queue, because that hides how big the load really is. The book is written for operations teams, but the idea transfers to any product team.

A simple version for a team of four to eight:

The interrupt rotation

  • Who: one person at a time, rotating weekly. Include everyone who can handle the work, including senior engineers.
  • Does: first response to incidents; triages every incoming request and question; fixes small things directly; turns bigger things into backlog items with enough detail for planning; keeps the unplanned-work log up to date.
  • Doesn't: take sprint items on the critical path. They can pick up small, non-urgent backlog items when the queue is quiet, but nothing the sprint goal depends on.
  • Escalates: to the manager or tech lead when an incident needs more than one person, or when the queue is more than one person can handle.
  • Handover: a 15-minute handover at the start of the rotation: open items, anything in progress, anything recurring.

Two things make it work. First, people outside the team need to know where to go: update the team channel's description or pin a message, and redirect direct messages ("Thanks, I've passed it to Priya, who's on interrupts this week"). Second, it has to be a real job. Count that person's capacity as interrupt time in planning, and don't judge their week by sprint items closed.

If the team is too small to give up a whole person, rotate by day or half-day. Either way, each person starts the day knowing whether they're doing focused work or handling interrupts.

Decide quickly what waits

The person on interrupts needs rules, or they'll end up negotiating every request from scratch. A short decision table, agreed with your product owner in advance, settles most cases:

Situation What happens Who decides
Production incident, data loss or security issue Handle now; pull in more people if needed Person on interrupts, escalating to the manager
User-facing bug with no workaround Fix this sprint from the unplanned budget Person on interrupts
Bug with a workaround, or small internal request Backlog, ordered at the next refinement Product owner
New feature or "small" enhancement Backlog; never added mid-sprint by the team alone Product owner
Question Answer, then add the answer to the docs if it's the second time Person on interrupts

The row that causes most trouble is the "small" feature. The Scrum Guide says that during a sprint, no changes are made that would endanger the sprint goal; new requests belong in the product backlog, which the product owner orders. That doesn't mean the answer is always "next sprint". It means the trade-off is made openly, by the person who owns priorities, instead of absorbed silently by whoever saw the message.

What to say when you're asked to squeeze something in

Most stakeholders will accept a "not now" if it comes with a reason, a time and an option. A script you can adapt:

"Thanks, that's clear. The team's working toward [sprint goal] until [date], and adding this now would put that at risk. I've added it to the backlog and flagged it for [product owner], who'll look at it for the sprint starting [date]. If it really can't wait, the question for [product owner] is what we drop to make room. Do you want me to set that up?"

The last sentence matters. It doesn't refuse; it turns the request into a trade-off. Sometimes the answer is yes, it's more important than what's planned, and something comes out. That's a legitimate decision, and a better one than the team quietly working evenings to fit both in.

When the budget runs out

Some sprints will blow through the budget, usually because of a big incident. When that happens:

  1. Swap, don't stack. Unplanned work that goes over budget pushes sprint items out; it doesn't get added on top. Tell the product owner which items are now at risk as soon as you know, not at the review.
  2. Protect the sprint goal before the item list. The goal is the commitment; the items are a forecast. Drop items that don't serve the goal first.
  3. Bring the overrun to the retrospective. One bad sprint is noise. Two or three in a row means your budget is wrong or the source of the interrupts needs fixing.

Shrink the source, not just the impact

Rotation and budgets make unplanned work manageable. They don't make it smaller. For that, use the log.

Every few sprints, sort the log by cause and look at the top three. The SRE chapter recommends regular reviews of tickets and pages to find root causes, and the habit works for product teams too. Recurring patterns usually point to a fix within the team's reach: a flaky job that pages every week, a report sales asks for by hand every month, an API nobody outside the team understands.

Take the biggest one into the retrospective and turn it into a single, finishable action, as described in Retrospectives that change something by the next sprint. Check the log a few sprints later. Seeing 35 hours a sprint fall to 25 convinces the team and your manager, and frees capacity without hiring anyone.

Your first step this week

You don't need to roll all of this out at once. This week:

  1. Add an unplanned tag to your tracker and ask the team to use it on anything over 30 minutes that wasn't in the plan, with a category and the hours spent.
  2. Pick who handles interrupts for the next sprint, and tell the people who usually message your team directly where requests should go now.
  3. Book 15 minutes with your product owner to agree the decision table above, especially the row about "small" features.

After two sprints, you'll have a real number for planning. Then ask yourself one question: if you showed the log to your stakeholders, would they be surprised by how much of the team's time goes to work nobody planned? If the answer is yes, that number belongs in your next status update.

Working toward PTMP Practitioner?

The free study guide covers every competency area on the PTMP exam, with practical examples.

More from the blog

Practical articles for technical managers at every stage. All posts

  1. Succession planning for your managers before anyone resigns

    A one-page succession plan for the critical roles under you: who could step in, when they'd be ready, and what each person does this quarter to close the gap.

  2. Making a technical debt case your stakeholders will fund

    Turn a vague request for tech debt time into a one-page case with a measured cost, a price, a payback period and options your stakeholders can choose between.

  3. Raising a performance problem while it's still small

    How to prepare and hold the first conversation about an engineer's slipping performance: a script, likely causes, tricky reactions and the written follow-up.