---
title: "Why software estimates are unreliable | QMates"
source: https://qmates.tech/en/services/engineering-transformation/software-delivery-assessment/unreliable-software-estimates
description: "Software estimates that are always wrong and dates that slip every sprint. Tangled code, oversized work, waiting between teams and a flow nobody measures."
updated: 2026-11-21
---

Unreliable software estimates

# Dates keep slipping and nobody can say when it will be ready

A date slips, so next time whoever asked multiplies the estimate to be safe, and the work grows to fill the margin. After a few sprints the estimates are unreliable, and neither the people who give them nor the people who receive them trust them any more. Underneath you find a system more tangled than it looks, work too big to estimate, waiting between teams and a flow nobody has measured.

## Does this sound familiar?

-   Every sprint ends with a few surprises, and the work that was ‘nearly done’ slides into the next one.
-   People who ask for a date multiply the team’s estimate ‘just in case’, and the team has got used to that margin.
-   A change that looked small touches parts of the system nobody had counted on.
-   The business has stopped asking for dates, or asks for them knowing they won’t be met.
-   When a date is missed, nobody can say where the work stopped between idea and release.

## Why are estimates so often wrong?

The useful question is about the system. How predictable is the work you are estimating, and how much do you really know about it? The causes below make it unpredictable, and the last one stops you noticing, because without measures of flow every wrong estimate looks like the fault of the person who gave it.

1.  ### The system is more tangled than it looks

    When parts of the code depend on each other without clear boundaries, every change drags others along. The team estimates the part it can see, while the time goes on the part it can’t.

    Feeds into

    -   
    -   

    [

    Handled by

    Legacy Code Refactoring](https://qmates.tech/en/services/engineering-transformation/legacy-code-refactoring)

2.  ### Estimates are made on work that is too big

    Big pieces of work hide more unknowns than their estimate shows, and you only find them at the end. Splitting the work and releasing it in small pieces shortens the time in which an estimate can go wrong.

    [

    Handled by

    DevOps, Continuous Delivery and Platform Engineering](https://qmates.tech/en/services/engineering-transformation/devops-continuous-delivery)

3.  ### Waiting between teams and urgent work that interrupts

    Part of the time goes between one hand-off and the next, when work passes to another team and waits its turn. Urgent requests that arrive mid-sprint, with nothing protecting the plan, add to it.

    Feeds into

    -   

    [

    Handled by

    Engineering Team Design](https://qmates.tech/en/services/engineering-transformation/engineering-team-design)

4.  ### The work that ships is never counted

    Without data on how much work really ships and where it stops, every estimate starts from zero. With the time that past work took from start to finish (the cycle time) you can give forecast ranges, which say how safe a date is.

    [

    Handled by

    Software Delivery Assessment](https://qmates.tech/en/services/engineering-transformation/software-delivery-assessment)

- [The service that tackles it](https://qmates.tech/en/services/engineering-transformation/software-delivery-assessment)

## What changes when delivery becomes predictable

The team uses estimates as a working tool. Dates are updated when something changes, and the people who receive one know how far to trust it.

When a date moves, the first question in the meeting is what to drop or bring forward, and business planning rests on what the team can forecast. Trust between the people who build and the people who decide is rebuilt on what actually ships.
