Introduction
A few years ago, I discovered a new way of projecting the future of a project, one that was more pragmatic and didn't involve divination. I'm now keen to train people in this, especially as it applies to much broader fields than just projecting the landing of a project.
My early days in IT were marked by project management in the industrial era. We used to spend a lot of time imagining the cost of implementing this or that functionality, or on a broader scale projects. To do this, we relied on what we knew at the time and on abacuses from the past or from other contexts. Of course, since every context was different and our current knowledge was often limited in relation to all the risks that would arise, we never got it right and very often got it very wrong. This was often painful, and did nothing to win the trust of our stakeholders.
An open path to a radical change
Then the Agile movement caught my eye. I learned a lot of interesting things and rediscovered some of the values I hold dear in this new paradigm. There were only a handful of us who were enthusiastic about it, and we made mistakes because of this newness. The mistake we made, and that many are still making today, lies in those landing estimates and projections. Certainly blinded by the much more collaborative aspect of planning poker sessions, we based our estimates on the Fibonacci sequence. From past performance (velocity), we projected the likely arrival of deliveries, taking the average of our velocity and adding a cone of uncertainty, because there was still a lot of uncertainty!
And once again, we didn't get it right, or else the cone of uncertainty had to be very wide, but far too wide for the stakeholders to accept.
Fortunately, a number of things will open my eyes to our stupidity and ignorance in trying to deterministically project things that were happening in a volatile, uncertain, complex and ambiguous world.
An approach which, once known and mastered, proves simpler and less time-consuming than previous ones, but above all much more accurate and much more conducive to having constructive conversations with stakeholders who have a legitimate need to know when and how much.
A real-life application in a field other than IT
It was during one of my training courses, entitled ‘Applying metrics for predictability’ AMP, that a member of a department responsible for steering the objectives and financial projects contractualization of the company X, accompanied by the product owner of a DATA solution equipping them, discovered this approach.
They spent two days learning and deconstructing ways of projecting the unknown, notably by avoiding the flaw of the average. This terrible trap was not adequately avoided by the statistics of the industrial era in today's uncertain world and in areas of high uncertainty. Indeed, these statistics are based on particular distributions (normal, binomial, poissonian, geometric, etc.), but in the majority of cases, your system doesn't generate any particular distribution, other than a chaotic one on which statistical laws no longer work.
To deal with this problem, a tool called Monte Carlo simulation has been created. With this type of simulation, it is possible to envisage hundreds of thousands or millions of possible futures, based on a real-life history.
You literally get Doctor Strange's powers - how classy is that?
This projection, if your system is stable, will enable you to obtain a distribution offering a highly accurate prediction. You can then select the level of risk appropriate to your context. Do you want 90%, 80% or 70% certainty?
The method is so simple that it is possible (and even necessary) to replay these simulations very frequently. Thanks to the new real-life data acquired, it is possible to refine the projection and identify any changes in trajectory that might require the activation of alert procedures.
This is extremely virtuous, as meetings are only held if a problem is detected, rather than on a regular basis, thus avoiding unnecessary meetings and enabling decisions to be taken more quickly. With this model, we have factual elements that point the finger at the issues to be addressed and encourage us to take decisions more quickly, with that clear sense of urgency.
The team in charge of steering therefore used their historical data to simulate the activity forecast for the end of 2024.
They also did exactly the same exercise they've been doing for decades to predict this landing.
With the benefit of hindsight, they now have the actual results obtained, enabling them to compare the accuracy of the various predictions made using the different methods.
The use case in action
Using the traditional method, which involves using a six-month moving average and making adjustments in comparison with years N-2 and N-1, the number of projected projects was exactly 28.
Given that they used an average, if the level of uncertainty were taken into account, the result would be a probability of 50% (in reality, this percentage is even lower, but let's not go into the details).
So, basically, the steering team was flipping a coin on this prediction of the future, with a very precise number and no room for other possibilities.
Using the knowledge acquired during the AMP training and thus this Monte Carlo simulation, they ran several simulations over different historical periods. For each simulation, they ran 670,000 draws.
Taking into account the different contexts and analyzing the one that appeared to be most equivalent to that of year-end 2024, they chose to announce year-end 2024 projections on the basis of the 2021 context.
There was therefore a 50% chance that 40 or more new projects would be contracted, a 70% chance that they would contract 32 or more projects, and an 85% chance that they would contract 24 or more projects.
Obviously, they had much more detailed information at their disposal, enabling them to select the probability exactly as they wished.
In reality, 31 projects were contracted, while the Monte Carlo simulation gave a 74% chance of this happening.
The 2022 sample was the closest to reality, with an 87% chance of this happening. It might have been wiser to rely on this history, but that's the complexity of our times: it's impossible to predict with certainty that what we're doing or taking is the right solution.
The old projection technique was less accurate, since it announced exactly 28 projects. It was a precise projection, but in our VUCA world, it's impossible to be precise; we have to aim for accuracy.
That's the power of this technique: it provides additional information about the probability of an event and a range of future possibilities.
If the 15% risk of completing only 24 projects poses a problem, then the simulation result can be used to sound the alarm and act accordingly. If this is not the case, then “business as usual” - no need to get together and make a big fuss.
As Sylvain from the steering team so rightly puts it: “The advantage is that we go from the dowsing rod to the wave scanner with 3D imaging of the subsoil.”
And all this for the same or even less effort and energy than before.
So why deprive yourself?
And that’s not all
System stability analysis is another advantage of this approach. With exactly the same data, it is possible to produce various analysis graphs, such as a cumulative flow chart, a scatter plot of completed projects, an age chart of current projects, or a process behavior chart.
All these additional elements enable further analysis and alert generation, at the right moment and based on factual information, rather than on feelings that are often too optimistic, leading to bad decisions or wasted time (see my article on this subject – in French): https://www.linkedin.com/pulse/pourquoi-loptimisme-peut-amener-%25C3%25A0-l%25C3%25A9chec-de-votre-produit-coignard-ooavc/?trackingId=dl9dbBJxQDy77UdTVGZLWg%3D%3D
By analyzing these other graphs, we could see that the system on which the Monte Carlo simulations were performed is unstable, which makes the predictions less accurate. You will note, however, that the predictions are better than those obtained with the traditional technique. All this information is a rich resource for the service, which can refer to it to address the appropriate topics and improve performance.
Finally, using the same method, the team has carried out simulations not on the number of projects for a given date, but on the amounts that could be committed in the future. Another rich source of information that will enable them to discuss and act with more factual elements and a perception of the future risks involved.
Conclusion
In conclusion, AMP ProKanban classes represents a unique opportunity to transform your approach to piloting in an ever-changing world. By integrating advanced system analysis techniques and tools such as Monte Carlo simulation, you won't just be predicting the future, but learning to navigate uncertainty with confidence and precision.
Imagine being able to anticipate outcomes with unrivalled accuracy, based on real data and robust analysis, rather than rough estimates. With an AMP training course, you'll have the tools you need to engage in constructive conversations with your stakeholders, providing them with clear projections and risk levels tailored to your specific contexts.
Don't leave the future of your projects to chance. Joining the AMP training course means choosing to move from the “divining rod” to a true “3D imaging” of your environment and your process. It's an opportunity to develop your skills, improve your team's performance and make informed decisions that will propel your projects to success.
Don't wait any longer to make the leap to more effective and relevant project management. Register today for AMP training course and become a key player in the transformation of your team and your organization!
Ajouter un commentaire
Commentaires