On Cycle Time for single item forecasting

There’s always a shiver running through the spine of any professional (and this is true both in and outside of tech I assume) when people start talking about estimations, deadlines and “give me a date”.

I know this well, I am no stranger to this shivering. Anytime a new item is added to the pile of things to be done, the question comes naturally and you, as the owner of the item presumably, will have to respond.

Now, we acknowledge right here and now that there’s many methods to come up with a guestimate (a guess of estimate) of sort, but as much as we try to pinpoint a specific date, we’re very much trying to tackle entropy with a deterministic model.

Entropy on its own will not bow to determinism. There is no algorithmic way to bound a new point on an X and Y axis with deterministic precision in a real world scenario. That’s it.

The work we do every day, we do in the real world and as such it is bound to the natural entropy of the real world. There’s a painter randomness in the strokes, but the strokes do work towards a common objective, the painting.

The question comes natural then. How can we do any estimations if that is the case?

The answer lies in embracing the non deterministic nature of the cycle of delivery items whilst acknowledging that through the laws of big numbers there will be some opportunity of convergence.

For the rest of the essay I will solely focus on delivery items from a software perspective. A delivery item can be thought of as a tech story or an epic.

Take a scatterplot, with its X axis measuring time and the Y axis measuring cycle time

From the graph in front of us we can gather the following informations:

• most items have been completed within a certain period of time
• those items that took longer to be completed are on the upper bound of the Y axis
• we have no way to know where the next completed item might be plotted (we can make educated guesses, more on this)

If we accept that we cannot deterministically plot points until we have all the informations for these work items being delivered, then what we can do is utilise those same completed items to build up a very simple forecast.

We are shifting our mindset by not answering “it’s going to be ready by this day” but rather “we have a % chance that all items are going to be delivered within Y days”

Specifically for the presented graph, we have 50% chance that an item might be delivered in 7 days, 85% chance that an item is delivered within 15 days and 95% chance that an item is completed within 22 days.

How can I tell? Because the graph has percentile lines drawn based on where the items are placed. This means that 85% of the delivery items are below the 15 days threshold. It also means that we complete 85% of our work items within 15 days statistically.

The mindset shift changes everything. We become owners of this decision based on real world data which presents itself as a series of probabilities which we can use to make informed decisions.