Your forecasting model is giving inconsistent results. How will you navigate through this uncertainty?

Forecasting is there to tell a story in your business and provide you with a guide of what's possible. If your forecast is consistently inaccurate you should look to the individual components to understand where the variable comes in. A business should be regularly reviewing the forecast and making the necessary adjustments to real time trends to ensure for accuracy.

Forecast accuracy is measured how close you are to your model. For time series ARIMA is a good method for your model and two of the main components in the analysis are Trend and Seasonality. To understand deviation from the model need to investigate the root cause for deviation in the seasonality and the trend. For example if we sell ice-cream on the beach and our forecast is not accurate, could it be that the weather is not good enough for going to the beach? Same example can be applied for many other scenarios.

To address inconsistent forecasting results, thoroughly scrutinize your data for completeness and accuracy. Identify any anomalies, and conduct a comprehensive data audit to rectify issues and stabilize your results.

1. Review your data once, twice/ as many times possible to eliminate inconsistencies. 2. Make sure that the data gathered is reflective of what is needed to make an accurate forecasting model. 3. The final results/ figures should be more consistent

Begin by assessing the quality of your data. Look for inconsistencies, errors, missing values, and outliers. Use data profiling techniques to understand the distribution and characteristics of your data. To further validate your data, cross-verify your data with multiple sources to ensure its accuracy and reliability. If discrepancies are found between different sources, investigate the reasons behind them and determine which source is the most trustworthy. Finally, compare historical data with actual outcomes to validate the accuracy of your data. Look for patterns, trends, and correlations to ensure that the data accurately reflects past performance.

In such a case, I primarily focus on 2 things. First, I review all sources of data and review how I tabulate it in the software or Excel. The second one, I review the assumptions that I made during forecasting. In 90+% cases, my problem will get solved.

Evaluate Data Quality: Review Model Assumptions and Parameters: Analyze the model structure: Implement Cross-Validation: Incorporate Robustness Checks:

Once you’ve confirmed the integrity of your data, the next step is to scrutinize the forecasting model itself. This involves a thorough review of the algorithms and formulas used to generate outputs. Look for any errors in calculations or logical flaws that might be contributing to inconsistencies. Additionally, assess any recent modifications to the model, as even minor tweaks can significantly affect results. If needed, consider reverting to an earlier, stable version of the model to analyze which changes may have introduced instability. This comprehensive review can help you pinpoint and address the root causes of forecasting discrepancies.

My advice is to follow 4 specific steps: Key Algorithm and Formulas Examination: Pay attention to math expressions, logic, and any custom functions used in the model. Error and Logic Checks: Scrutinize calculations step by step. Ensure there are no computational errors or logical flaws. Recent Changes Audit: Investigate any recent modifications to the model. Even seemingly minor adjustments can impact results. Reversion Analysis: Compare the stable version with the current one.

Zero base budgeting after a time gap can also be used to justify all the logics, assumptions and parameters of an existing budget and forecast methodology used by an organisation. Zero base budget may work well as a cost control tool where the management intends to do so. At the same time it may give a fresh perspective too.

Document the assumptions. They should be a series of switches. Someone (not in finance) will come up with a "must have" scenario so the easier it is the change, the faster you can see the results.

Inconsistency might be due to assumptions, we need to understand the business model first and consider the cyclicality and future growth after that scenario, and sensitivity testing might help you to achieve your targeted results.

Here's how I'd phrase it: consult experts before putting in data, or assumptions. An outside party will be able to give perspective, but internal experts know what drives what. I've always benefitted by understanding what's going on in the Engine Room...

This can really get in the way. Periodic upgrading is fine. Changing rapidly and often leads to a series of potentially vastly different reports, which are hard to compare and contrast.

Inconsistent forecasts are a sign to refine your model. Regularly incorporate new data and feedback to improve its accuracy and resilience against future uncertainties. This continuous improvement approach is key to ensuring your model stays reliable over time.

A finance model's results and reports are as good as your assumptions and the drivers it's based upon. At one of the billion-dollar companies I worked for, these are the three methods we've used to continue making the model better, and the organization perfected the model over a period of time:- 1. Test assumptions using historical data and test the model against them to see if the outputs are consistent with the expected results. 2. Let cross-functional team leaders and c-suite challenge the assumptions, and assumptions are defended using pragmatic methods and realistic data to support them. 3. Validate assumptions made in a project after one to three years, based on the ROI payback promises, and continue making corrections. #Rahar

Foster collaboration between different departments within your organisation. Cross-functional input can lead to more comprehensive and accurate forecasts by incorporating diverse perspectives.

I've noticed that in some organizations, even if they are using IT systems for enterprise performance management, integrated business planning, extended financial planning and analysis (FP&A), and using these systems for the annual budget, monthly forecasts, quarterly business reviews, they are still using old excel models for long term strategic planning or analyzing their investments or still keeping certain financial modeling work outside of the system. Most of the time, it's the need for more education and understanding of the capabilities of the EPM Systems that could incorporate complex models, and that FP&A team could use Corporate Financial Performance Management Systems, which do not corrupt as an #ExcelModel would. #Rahar