What is a Good Estimate?

By Ko Ito

November 6, 2024

In this article, I would like to consider what a good estimate is using the following three scenarios as examples.

  1. A task that could previously have been done in 3 days is estimated to take 5 days, and is completed in 5 days.
  2. Two members of the team estimated 8 and 12 days, respectively, but scheduled the task to take 10 days.
  3. A task that had previously taken 9 days is estimated to take 8 days, and is completed in 9 days.

Which do you think is the best estimate? It depends on the work being done, but this time let’s assume intellectual work like software development rather than labor work to make it easier to think. Let’s examine each one together.

There is No Correct Answer to the Value of Estimate
The first scenario was completed in 5 days, so can we say that this is a correct estimate? If the work had been estimated in 6 days, the quality of the work might have been improved. Or, if the work had been estimated in 4 days, it might have been possible to finish it in 4 days somehow. There is no correct answer to the value of estimate because we cannot verify it even after getting actual values.

The estimated time belongs to the worker, and it is usually all used up. Of course, this doesn’t mean finishing the work in three days and doing nothing for the remaining two days. When it comes to intellectual work, it’s difficult to draw a clear definition of done and say that this is enough, and there are many things that can be done to improve the quality of the work.

The Important Thing about Estimates is to Eliminate Discrepancies
For the second scenario, it seems fair and good since the middle value is taken out of respect for both opinions. However, this does not utilize the perspectives of both people at all. An estimate is the allocation of values (time or cost) to work. As mentioned above, there is no correct answer for the values, so it is impossible to refine them. To make a good estimate, you can minimize the discrepancy in interpretation of the work content written in natural language such as Japanese or English. It’s quite simple. Just ask each person, “Why did you think so?”. Then you can notice the discrepancy in the interpretation of the scope of the work between the two people.

Since all estimates are inferences from similar past experiences, you can learn from others by listening to specific past cases, and you can also consult with your team and the customer to decide which interpretation is appropriate in this case. As a result, a clearer definition of done can be achieved, minimizing the risk of scope creep.

Planning Poker, used in Agile, is a tool exactly for this purpose. It is often explained as an estimation tool, but more accurately, it is a tool for resolving discrepancies. The purpose of doing Planning Poker is to provide an opportunity for dialogue to confirm members’ tacit knowledge, such as underlying assumptions and constraints.

The Purpose of an Estimate
For the third scenario, it is not a good estimate in terms of taking 9 days instead of the estimated 8 days, which will cause some trouble to later processes, and if it is a task on the critical path, it will affect the overall project schedule. However, considering the purpose of estimates, it is not necessarily a bad one. Peter Drucker has the following to say about action plans:

 The action plan is a statement of intentions rather than a commitment. It should be revised often, because every success and failure create new opportunities. An action plan is also necessary as a standard for time management, and serves as a guideline for how to use time.

A statement of intentions is like a target time when participating in a marathon, and it leads to increased motivation. In the case of a marathon, it depends on your physical condition on the day, but, if possible, you should aim to beat your personal best. So, in this case, it may have been appropriate to set the time to eight days to improve one’s personal best.

Bottom-up Estimation as a Baseline
Bottom-up estimation is necessary to get what Drucker calls a guideline for how to use time. Setting detailed lap time targets, not just the goal time, allows for early detection and early response. For example, if you aim to complete the race in 4 hours and 15 minutes at a pace of 1 hour for 10 kilometers, that would be the baseline. There is no problem if you are under an hour when you pass the 10-kilometer mark, but if you realize at that point that you are two minutes behind, you can make up two minutes in the remaining 32 kilometers. However, if you realize at the 40km point that you have only 5 minutes left until your target time, it’s too late to make up. The same idea applies to costs.

The purpose of bottom-up estimation is early detection and early response as mentioned above. However, many people believe that bottom-up estimation is used to improve the accuracy of estimates. In fact, when I ask training participants whether they think top-down or bottom-up is more accurate, about 80% answer bottom-up. As mentioned above, there is no correct answer in estimation, so it is impossible to verify which is more accurate.

So why do people have the impression that bottom-up is more accurate? This is coming from the difference in their purposes. Top-down is used for rough estimates before detailed specifications are decided, while bottom-up can only be implemented after the WBS has been defined, which is why people have the impression that bottom-up is more accurate.

Examples of Rough Estimate
Whether top-down or bottom-up, humans can only make inferences from similar past experiences. However, in certain fields, it is possible to come up with an overall cost by parametric estimation.

For example, in a construction project, even if there are no blueprints or specifications, the cost can be estimated by calculating the cost per square based on the actual cost of past projects with similar regions, uses, scales, structures, grades, etc., and multiplying this by the floor area. At this time, fluctuations in material and labor costs are taken into consideration, but even so, some gaps are unavoidable because of differences in the shape of the site and building and differences in ground conditions. Similarly, construction costs can be estimated by setting unit prices based on the main functions of the building, such as “number of dwelling units” in an apartment, “number of beds” in a hospital, “number of seats” in a theater, and “number of rooms” in a hotel.

Lastly, I would like to introduce an interesting example of an estimate. Maeda Corporation is a major Japanese construction company with a track record of constructing many dams, tunnels, bridges, and the like. Volunteer employees of this company launched a project to calculate rough estimates of construction costs and periods assuming the actual construction of civil engineering facilities that appear in science fiction and anime works, and publish them on its website. For example, the construction period for the hangar of the “Mazinger Z”, which is 1970s Japanese anime, will be 6 years and 5 months, and the total construction cost is 7.2 billion yen (1.1B for excavation and 6.1B for machinery and equipment). This project was made into a movie in 2020, and it became an opportunity for the estimates in the civil engineering and construction industry to be widely known to the public.

This creative project highlights the versatility of estimating techniques, demonstrating that they can bring imagination to life while showcasing the practical value of estimates across all types of projects.

Ko Ito

Freelance Trainer, Translator and Course Developer
Consultant and Trainer, International Institute for Learning

Ko’s experience spans over 15 years, and he has provided various online and offline trainings in Project Management, Business Analysis, Leadership, and Agile. Prior to teaching, Ko worked with several American IT companies including DEC, HP, and Intel.

Ko also teaches courses in several schools including the National Institute of Technology, Keio University, and Ishikawa IT Center Business School. Additionally, he has worked as a trainer at Botswana Public Service College in Africa.

Ko earned a Bachelor of Science in Mathematics and MBA from Waseda University in Tokyo. He finished his Doctoral Program at the Tokyo Institute of Technology, Graduate School of Innovation Management. He is the first Certified Business Analysis Professional (CBAP®) in Japan.

Visit Ko’s social media links to learn more.
Facebook: facebook.com/ko.ito2
LinkedIn: linkedin.com/in/ko-ito-japan
Twitter: twitter.com/ko_ito

Scroll to Top