One of the key-metrics in every business that defines success is SLA, or – Service Level Agreements. Metrics, or values that SLA defines are based off of the business it supports, and it may be anything that is needed: sales volume, number of calls, chats, number of parts used, and medication sold, mortgages, and similar – within the defined time-frame. We’ve done a brief introduction to Data and SLA forecasting in this article, so check it out, if you’re interested.

Therefore, besides the certain metrics that you will choose, that are of importance to your business (or the one of your clients) – time-frame is equally important aspect. Work orders, and contracts, do consist of those two things, usually, and they do provide an insight into whether you’re reaching the agreed target level, or are you below, or over-delivering. Let’s analyze one simple chart below, which may be looked as an SLA, with certain volume, over certain period of time:

On the ‘X-axis’ we have a time-frame for which we are looking the results, while on the ‘Y-axis’ we have the values. Just glancing through the data flow, over this period of time, we overall see the nice, even elegant, and constant trend…except for a certain spike, or ‘anomaly’ that has happened around June 15th. So – this anomaly is not a trend, but it is ‘something’ that has happened around that day, and caused the spike in data.

When looking into the SLA, although the data flow, and success rates are important, on the 2nd place do come the anomalies, as they will help us to understand as to what has happened on particular date (in this example) that has caused our results either to spike sky-high, or to go down-low. With the help of Machine Learning, and the needed algorithms, this data can be automatically analyzed, certain volumes extracted, and calculated, so that we can understand as to why something has happened, what we need to do to repeat it again, or what we shouldn’t do, in case we want to avoid this from happening again.

Machine Learning, for this particular use-case, has advantages over devoted human resources, as there is less space for human error, and this comes down to math, calculations, and predictions – less space for errors, more accurate the results will be and, over the course of time, algorithms can adjust to the needed.


Once you’ve gathered the data, next step would be to send it out to the in-house experts, or to the company of your choosing that can compile the data, and ‘make sense’ out of it, with the needed analysis, and provided algorithms. Here, in this phase, it is important to have a constant, two-way communication between the stakeholders in this project, since it provides the foundation of the project. On one end, you will have the business experts – the ones who understand what the data means, while on the other end you will have a company, or the department, that will find the correlation between the data points/documents provided.

In this phase it is critical for the communication flow to be open, and constant (if need be), since usually both parties involved have a lot of questions, and need better clarity, and understanding, of what the purpose of the data is, and what the desired outcome needs to be, so that the goals can be defined.

On the end of the company that provides the data, it would be good to have representatives from the business end (somebody who is involved in daily operations) and more technical person also (somebody who understand the overall technologies used), so they can respond to the needed questions that department, or company, tasked with the Machine Learning implementation might have.

Depending of the amount of data provided, and importance of it – initial data analysis may take some time, and that is something that is being agreed upon, usually, once the data is presented, and the overall goal of the project is known to the company that does the analysis, and Machine Learning implementation


Looking back at the chart above – once we’re done with the anomaly analysis, the rest of the chart shows us a good, constant, flow, and we can easily see the ‘ups and downs’ that are almost equal in repetition, over the certain periods of time (in the case presented – these are, obviously weeks, which can be seen as per dates on the ‘X-axis’). Besides slight variations, and if we presume that the value ‘5’ on our ‘Y-axis’ may be our contracted target levels, we can see that, in the majority of the cases we are reaching the needed value, over-delivering in certain periods of time, and on occasional days – we are slightly below the target levels. Here, when it comes to SLA negotiation, and contracts, it is (usually) defined that the target levels of SLA’s, for a certain period of time, need to hit (for example) at least 80% success rates which, in this case, can be done by looking through the average values: if, at the first half of the week you are at a 100% of SLA, and in the second part you are at 70% – your average value for that week is at 85% and, if your weekly contract is at 80% – you’re even above it.

Sure – it is, usually, better to over-deliver, rather than to be below the contracted line, however – constant over-delivering does bring up the most frequent down-side of all: during the next contract renegotiation, your target levels, if constantly over delivering, may be easily re-adjusted, and the bar raised higher, bringing in the unrealistic expectations. This is one of the parts where, in addition to analysis, Machine Learning can help you – with predictions, and understanding if anything needs to be readjusted in real-time, rather than as a ‘consequence’, or reaction – once it’s too late.

Overall, when it comes to reaching the targeted levels, staying above, or analyzing anomalies, with the help of Machine Learning will provide you, your company, and your clients for the results to be right where they need to be, and also to have better insight into the future of your business, so you can schedule, and plan, accordingly.


Machine Learning is bringing great value to the companies, and it is becoming more and more understood, and popular (of course, we do need to step away from the ‘Terminator’ movies scenarios here, please ☺). Not necessarily Machine Learning will always be the proper technology that will bring the solution for your business – this part, or the idea, should be left to the people that are experts in that field, and based off of the initial analysis, perhaps a different solution will be implemented.

Perhaps, it will be another piece of software that is not even related to AI, or Machine Learning. However, if you do find certain things in this article, that drive you towards Machine Learning solution, or even if it made you more curious about the benefits of it, don’t hesitate to contact us, and we’ll be more than happy to chat with you.

Thank you for reading!

Author: Marko Djapic

Read more posts from the author at Rubik’s Code.