Teams that ship software keep asking the same questions about time. How long will this change take? Why did the last one take longer? Should we trust the number we gave? The answers are rarely complicated, but they are scattered across habits that few teams write down. This page collects seven of the most common questions in one place, each answered directly, with one consistent theme: a forecast is only a guess until you compare it with what happened.
1. Why do delivery estimates usually come out too short?
Because people picture the path where everything works. An estimate is formed from the visible steps, such as writing the change, testing it and releasing it. It leaves out the invisible ones: a flaky test that needs rerunning, an approval that arrives a day late, a dependency that needs updating first. Each omitted step is small, yet they accumulate.
There is a second cause. Estimates are often spoken aloud in meetings where a short number is more comfortable than a long one. The result is a habit of optimism, and it can be measured. Record the guess, record the actual time, and divide. The typical ratio will show the size of your optimism.
2. What exactly should we record?
Four items per task are enough: the type of work, the estimate given before starting, the time actually spent from start to finish, and a one-line note about the main cause of any gap. Add the calculated ratio in a fifth column. Avoid recording more than this at first; extra fields tend to kill the habit.
Decide in advance how you will measure the actual. Some teams count only hands-on effort, others count elapsed time including waiting. Each has its uses, though they answer separate questions, so give them separate columns where possible.
One more practical point: record the entry at the moment the task closes, rather than on Friday afternoon. Recollection smooths over the awkward afternoon when a deployment stalled, and by Friday the detail is gone. A line typed in at closing time takes under a minute and preserves the reason for the gap while it is still fresh.
3. How many data points do we need before the ratio means anything?
Five or six tasks of the same type give a first impression. Ten give something you can plan with. Below five, treat the result as an observation rather than a rule. It is better to have a provisional factor than none, as long as everyone understands it is provisional.
Mixing different task types is a frequent mistake. A ratio averaged over small bug fixes, large migrations and documentation updates describes none of them well. Group the rows, then average within each group.
Be wary, too, of tasks that were split or merged along the way. If a ticket grew into two, add the hours of both back to the original estimate, otherwise the ratio will look flattering. Honest bookkeeping here matters more than neat columns.
4. Does a better estimate always mean a bigger number?
No. Most teams find that their error leans toward the optimistic side, so corrections usually raise estimates. But some teams, especially those burned by earlier overruns, overcorrect and begin to pad. Their ratios drop below one, and the correction is to trim. The target is a ratio near one, not a larger number for its own sake.
Watching the ratio over several months also shows whether improvements in tooling or process are real. If automation is supposed to speed up releases, the ratio for release tasks should move toward one, or the absolute hours should fall. If neither happens, the change may not be delivering what was hoped.
5. How do customer-facing goals fit in?
Estimates matter to customers indirectly. When delivery dates keep slipping, the people who rely on the product lose confidence, however good the features are. A discussion such as this article about how modern delivery practices influence customer satisfaction shows how closely speed, predictability and user trust are connected. Reliable estimates are the quiet mechanism underneath that connection: they let the team promise dates that it can keep.
If you publish roadmaps or release notes, consider applying your correction factors before you commit to a date. A slightly later date that is met is worth more than an earlier one that is missed.
6. How do we handle this in a team that resists change?
Start small and avoid framing it as monitoring. Invite people to track their own numbers for two weeks and then share only the patterns, not the individuals. People accept a measurement more readily when it is clearly aimed at improving plans rather than judging performance.
Culture shifts slowly, and a tracking habit is one of its small levers. If you are thinking about the wider picture, a practical piece on managing cultural change within DevOps work covers the human side of introducing new routines, and the advice there applies directly to a time-tracking habit as well. Pick one team, one task type and one month, and show the results to colleagues before expanding.
A final suggestion for resistant teams is to begin with the manager’s own estimates. When the person who sets deadlines shares a ratio first, the exercise feels like a shared experiment and not a request from above.
7. Can we apply this beyond software?
Yes. The same cycle of guessing, measuring and adjusting works for personal projects, household chores, travel plans and study. It is a general discipline, not a technical one, and teams often find that the habit spreads once a few members have tried it at home. Many organisations that work in education, research and community settings take a similar attitude toward looking back at results. One illustration is the Cotocus Foundation website, which represents a nonprofit organisation that connects technology, innovation and education, and lists sections for program, research, community and blog content.
A short worked example
Consider a platform team that logs twelve infrastructure change tasks over two months. For six routine configuration changes, the average ratio is 1.1. For six changes involving a third-party service, the average is 1.9. The team now knows that estimates for the second group should be nearly doubled, or that such work should be scoped with the vendor’s availability in mind. It also knows that its routine estimates are close to fine and need no padding.
That result would have been invisible without the table. The overall average, about 1.5, would have suggested a general increase of fifty percent, which would have inflated the easy work and still underserved the hard work. The grouping is where the insight lives.
Pitfalls to avoid
- Changing the estimate afterwards. If you adjust the guess once the work is underway, you erase the evidence of your original bias.
- Averaging across everything. A single team-wide figure conceals which kinds of work cause trouble.
- Using the ratio as a target. If people feel they must hit a ratio of one, they will bend the numbers. Present it as information.
Quick reference
- Write the estimate before work begins.
- Record actual effort and, if relevant, elapsed time.
- Group by task type before averaging.
- Apply factors to new estimates and watch the next results.
- Aim for a ratio near one, in either direction.
- Review the table with the team once a month.
These seven answers can be read in five minutes and put to use the same afternoon. Begin with the next task on your board, and in a few weeks you will have the first honest numbers about how your team really plans.









