top of page
  • Writer's pictureTash Joslin

Strategies for Working with Imperfect Data


MoJo's recent survey on Data Utilisation within UK SMEs identified a significant hurdle, with half of all responders stating the challenge of collecting accurate data.


This obstacle frequently dissuades SMEs from initiating their data-driven journey. Consequently, in the following discussion, we aim to present strategies to help businesses overcome this particular challenge.


In a perfect world, with unlimited budget and resources, every business will have perfect data, and the elusive single customer view will be at everyone’s fingertips with interactive dashboards flashing up as soon as you enter the offer.


However, even large enterprises struggle to make this a reality for a number of different reasons.

In the realm of data analytics, perfection is a rare and elusive creature. Real-world datasets are often messy, incomplete, or contain errors.


For many smaller businesses, grappling with imperfect data is not a choice but a reality.


The good news is that working with imperfect data is not a roadblock; it's an opportunity to refine your analytical thinking and develop resilience in decision-making.


Let's explore strategies for navigating the challenges posed by imperfect data and turning them into avenues for growth.


1. Embrace the Imperfection:


The first step in working with imperfect data is to acknowledge its existence. Rather than viewing it as a hindrance, see it as an inherent aspect of the data landscape. Recognising imperfection allows you and your analyst to approach it with a pragmatic mindset, understanding that the quest for flawless data is often impractical.


2. Understand the Limitations:


Knowing the limitations of your data is crucial, this is as important as a business leader or an analyst.


Be transparent and educated about what your data can and cannot tell you. Understand the context of how the data was collected and any inherent biases it might carry. Ensuring this is documented will be incredibly useful in the future especially when things change. This awareness helps in making informed decisions without overinterpreting the data.


A great example was an analyst reporting consistently that the most successful trading day was on a Monday but in reality, it was at the weekend so no one trusted their insight. In reality, the batch-upload of data was delayed by 24 hours but no one had informed the analyst.


3. Clean and Preprocess:


Imperfect data doesn't mean unusable data to show trends and patterns to aid decision-making. Being mindful of the previous point - Data can be cleaned and preprocessed to remove outliers, correct errors, and fill in missing values.


Numerous tools and techniques, ranging from simple spreadsheets to advanced algorithms, can assist in this process. Consider creating a data cleaning routine as a standard part of your analysis workflow.


4. Leverage Statistical Techniques:


There are many statistical methods that provide powerful tools for working with imperfect data. Machine learning and AI can support the delivery of these different methods to help fill in missing values, and techniques like regression analysis can provide insights even in the presence of noise.


It is important though to understand the statistical assumptions behind your chosen methods to make informed interpretations.


Robust validation processes can be put in place to assess the reliability of your findings. Cross-validation, sensitivity analysis, and other validation techniques help gauge the stability of your results under different conditions. This step is essential for building confidence in your analyses despite the imperfections in the data.


Support your analyst by using an external data analytics consultant can be helpful in starting on this journey and by putting in place some of these techniques.


5. Embracing Multiple Data Sources:


The development of Digital tools over the past ten years has been transformational for many SME’s. These tools are incredibly useful and often are providing valuable data back to the business.


This can be overwhelming though as now you have more data than you know what to do with and, therefore what should you be using to make a decision? Often one tool won’t give you all the answers but combining the insights will give you the bigger picture.


There can be frustration when reports and insight don’t 100% tie up, but this will require your analyst to be well versed in each data source (being mindful of limitations of each dataset) and be able to ‘tell the story’ of how the insights work together.


Give consideration when you are setting up the new tools as to how you want to report on the outcomes, using consistent definitions and terminology between tools will be beneficial going forward.


Integration between tools is becoming more and more achievable through Application Programming Interfaces (API’s), but they can be costly. It would be worth establishing the cost-benefit of doing it based on the return or if it’s something that you are dependent on and it is taking too long for someone to produce manually.


6. Communicate Uncertainty:


Transparency is key when working with imperfect data. Clearly communicate the uncertainties and limitations associated with your findings.


Trust is really important when working with data and it’s important to provide stakeholders with a realistic understanding of the data's imperfections to foster trust and ensure that decisions are made with a full awareness of the data's nuances.


Support your analysis development by understanding the imperfections, and context and establish the reason that data points have changed or have fluctuated significantly since the last report - especially before it goes to the Boardroom.


Conclusion:


Imperfect data is not the enemy of progress; it's an integral part of the analytical journey. By embracing imperfections, understanding limitations, and employing strategic techniques, businesses can extract valuable insights even from less-than-ideal datasets.


Working with imperfect data is not a compromise but an opportunity to develop resilience, enhance analytical skills, and make informed decisions in the face of uncertainty. Remember, the goal is not perfection but progress in the pursuit of better business outcomes.


If you need any support in this area, please get in touch at hello@wearemojo.co.uk



Comments


bottom of page