Ground Floor, Adamson House, Towers Business Park, Wilmslow Rd, Manchester M20 2YY

The six fundamentals of customer operations forecasting

Philip Stubbs outlines six essential elements of forecasting for a customer operation

In this article, I explore the six fundamental elements crucial for achieving accurate forecasting in customer operations such as contact centre, back office centre or field operation. These are: Data, Model, Process, People, Software and Charts. Each plays a vital role in improving predictions and enhancing decision-making.

As we go through each of the six elements, take a view of the capability within your organisation to determine opportunities to make improvements to forecast accuracy. The result can be significant commercial benefit from increased customer experience and revenue, reduced cost, increased performance and improved employee engagement.

1. Data
Data is extremely valuable for forecasting, the most important forecasting fundamental. Investing in data is crucial; in most cases I’d rather apply a simple model to accurate data than apply a more sophisticated model to unreliable data.

Prioritise data quality
Trace the data of the variable you are forecasting back to its original source to ensure accuracy and avoid misunderstandings. Also, make sure you are aware when customer volumes are constrained due to a channel being turned off or deprioritised, and ensure this is incorporated in your modelling. Stay updated on measurement changes and workflow adjustments, and keep engaged with data suppliers to maintain quality.

Clean and pre-process data
Keep your data clean to maintain reliability for forecasting. Use appropriate techniques to address outliers and missing values promptly to avoid misleading insights. Also, it may be necessary to assemble data manually from challenging sources.

Be up-to-date
Always work with the most current data. Using outdated information can ignore recent activity and lead to inaccurate forecasts. Incorporate the latest data to reflect current conditions and improve forecast relevance.

Understand variation in the forecast variable
Examine historical data to understand its variation. Invest time analysing historical volumes and document variation, consulting broadly to explain it. Such understanding helps you select the most appropriate forecasting model. Keeping a diary ensures reasons for variation are not forgotten.

Source explanative data
Collect reliable explanative data like company sales, promo activity, events, statement production, weather and social media activity – anything that might explain variation within customer operations volumes. Such data might be extremely useful in finding models with the best predictive ability. Maintain good relationships with data providers for reliable information.

Organise and make data accessible
Ensure your data is well-organised and easily accessible. Use consistent formats and tools that simplify data retrieval and integration into forecasting models. This improves efficiency and helps reduce error.

Example chart of historical weekly data – all variation should be understood before forecasting

2. Model
This is the method, equation or algorithm you use to turn the data and business insight into accurate forecasts.

Start simple!
Begin with a simple forecasting model, such as a weighted moving average, to establish a benchmark. Only add complexity if it improves accuracy above this baseline. Too often I have seen poorly-specified deployed complex models outperformed by simple models.

Study the causes of variation
Understanding variation in your forecast variable – resulting from seasonality, trend or external factors – is valuable. Proper treatment of these elements can help enhance model performance beyond the simple benchmark.

Incorporate external variables
Ensure that external variables have a useful relationship with the outcome you’re forecasting. Use tests to validate these relationships. Additionally, ensure that you can acquire reliable forecasts of these external variables for your predictions.

Don’t forget lags
In customer operations, external variables often have delayed effects. Test whether including lagged external variables improves accuracy.

Experiment with machine learning
Machine learning methods are increasingly accessible. Experiment with these techniques to see if they outperform conventional statistical methods.

Different time horizons
Models may perform differently based on the forecast horizon. Choose a model suited to the specific time frame of your operation.

Use a sensible selection method
Employ a robust error KPI and test models on separate data not used in model training to minimise overfitting. This helps ensure that the model performs well on new data.

Several scenarios
Different business perspectives on future marketing or technology impacts may require several forecasts to be created. Create scenarios based on different assumptions to test operational resilience and develop alternative strategies.

Find the right balance of complexity & interpretability
Within customer operations, particularly for long-term forecasts, the right balance between model complexity and interpretability is crucial. Often, simpler, interpretable models may be preferable to complex ones that are harder to adjust.

3. Process
A well-structured process ensures your volume forecasts are timely, accurate, and actionable. Here are key strategies for refining your forecasting process:

Reforecast frequently
Regular forecasting helps keep predictions aligned with the most recent customer behaviour and other changes & developments. If forecasts are done too infrequently, your operation may be caught off guard by issues that could have been foreseen and prepared for.

Use the latest data
If you’re forecasting on a Monday for decisions made on Thursday, you may miss critical insights that would have been available if you’d waited until later in the week. By forecasting as close to the time of action as possible, you ensure greater accuracy and provide resource planners with more reliable forecasts.

Ensure timely and accurate inputs
External inputs – like sales forecasts and marketing plans – must be both timely and reliable. Regularly assess the accuracy of these inputs, and establish service-level agreements (SLAs) to hold suppliers accountable for maintaining timely inputs.

Automate process where possible
Automating data cleaning, calculations and routine tasks minimises manual errors. It also frees up time to focus on more strategic elements of the forecasting process.

Study errors and continuously improve
Make error tracking a core part of your process. By analysing and sharing findings with stakeholders, you can uncover opportunities for improvement. Your forecasting process should evolve over time as you continuously refine and optimise based on lessons learned.

Apply judgement and welcome challenges
In customer operations, some upcoming events and changes may have unclear impacts. Incorporate expert judgment and operational insights to plan for these uncertainties. Track the performance of pre- and post-judgment forecasts to understand whether these judgments enhance accuracy.

Document the Process
Clear documentation ensures consistency and transparency across your team. When everyone understands the process, the chances for mistakes decrease.

4. People
Forecasting accuracy hinges on effective team dynamics. Here are some key factors to optimise the human side of forecasting:

Foster ownership and accountability
Clearly defined roles and responsibilities ensure individuals understand their responsibility for forecast process steps. Documentation and internal service levels can help with this.

Enhance collaboration and communication across teams
Clear and open lines of communication between teams (with eg Finance, Marketing & Operations) prevent misalignment and create an opportunity to achieve the best accuracy in demand drivers. Support from stakeholders enriches the forecasting process.

Invest in training & skills development
Equip forecasting analysts with a solid understanding of data interpretation, statistical techniques and machine learning methods. Skilled analysts can better address data inconsistencies and select the best model. Also, ensure that people skills such as communication, influencing and negotiating are covered in personal development plans.

Manage bias
Human beings often introduce biases into forecasts, whether it’s optimism from sales teams or caution from finance departments. Identify and address such biases in forecasts. By measuring inputs and understanding the impact of judgment-based amendments, teams can enhance accuracy.

Cultivate a supportive environment
Too often, I have seen forecast errors blamed for poor service levels, even when the root cause lies elsewhere. Create a culture that supports forecasting analysts rather than blaming them for errors. Focus on root causes and foster collaboration to learn from historical forecast errors.

Build trust and credibility
Trust between the forecasting analyst and other departments is essential. When stakeholders trust the forecast, they are more likely to rely on it for decision-making and provide valuable feedback. Credibility comes from consistently delivering reliable results and openly communicating any limitations or uncertainties.

5. Software
The right software is vital for achieving forecasting accuracy, enabling analysts to study datasets, identify & understand patterns, select the best model, make predictions and analyse errors.

WFM forecasting tools are integrated in Workforce Management systems, offering convenient forecasting capabilities that handle trend and seasonality without external software. The seamless integration allows forecasts to be generated efficiently as part of broader WFM functionality.

But variation and its causes differ significantly across businesses and worktypes, often meaning that a WFM’s single technique or limited methods rarely deliver the most accurate forecasts. This may negatively impact customer experience, revenue, cost, and employee satisfaction. But some organisations may perceive that the convenience of WFM forecasting outweighs the potential accuracy gains of more advanced models.

Excel is a ubiquitous, highly flexible tool for building and sharing custom forecasting models. It supports a range of statistical techniques, including multiple regression. With Excel, forecasting models are auditable and accessible to those wishing to understand them. Its charting capabilities allow leaders to review and critique forecasts, while quick adjustments to “what-if” analyses make it ideal for collaboration.

However, Excel struggles with large datasets and maintaining accuracy at scale. It also lacks the advanced modelling capabilities available elsewhere.

Python, R and SAS offer greater automation and a broad range of statistical and machine learning techniques, making them powerful forecasting options for the more challenging worktypes. They’re strong in handling complex data, and also selecting & testing forecasting models. All have community support.

These tools come with challenges, however, including steep learning curves and a need for specialised coding and statistical expertise. This can limit their accessibility for teams without strong analytical backgrounds. Having transitioned to Python this year, I’m very impressed with it, and I recommend it to analysts seeking to explore forecasting opportunities beyond the limits of Excel and WFMs.

6. Charts
The use of charts can significantly enhance forecasting accuracy. The insights they provide often go beyond what you can obtain from a table of numbers. Over the years I’ve seen how the introduction of charts to organisations has led to improved forecasting accuracy. Here’s how:

Understand Variation
Visualising historical data through charts provides a deeper understanding of past fluctuations and their causes, crucial to building more accurate forecasts. Charts make it easier to spot anomalies, seasonality, trends or recurring spikes linked to external factors. Recognising such variation helps you to specify and fine-tune the model that will deliver best accuracy.

Check forecasts
Comparing forecasts with recent actuals on a chart provides an immediate visual check to spot potential problems (eg see chart below). Ask yourself whether the forecasts look sensible given recent actuals, and are they aligned with what you expect to happen? This practice can help you catch potential mistakes early.

Build consensus with stakeholders
A visual representation of forecasts versus recent actuals serves as a powerful tool in discussions with stakeholders. By using charts alongside documented assumptions about future variation, you can facilitate clearer communication, build transparency, and ensure alignment across teams. Charts also allow you to demonstrate different scenarios (eg the uncertain outcome of automation initiatives), enabling more informed discussions on resource planning.

Learn from past errors
Forecast error is inevitable, but charts can be invaluable in helping you understand recent errors. By plotting forecasted values alongside actual outcomes, you can determine whether there are learnings that can help boost future accuracy. With this insight, you can refine your approach, minimise future error and continuously improve the operations planning process.

I’ve often been teased about my enthusiasm for charts, but I’ve seen time and time again how the right chart can vastly improve the value of forecasting to a business. By integrating the right charts into your forecasting processes, you can unlock significant benefits.

As I suggested in the introduction, a useful exercise is to review and determine your own organisation’s capability across these six forecasting fundamentals. Identify areas for improvement that can add value to your forecasting process.

If you would like assistance in identifying opportunities for improving operational forecasting, please contact us through this website.

========

Philip Stubbs is Partner of Atlantic Insight, and has over 25 years’ experience of improving performance within operational areas within a wide range of industries.

At Atlantic Insight, our mission is to help business operations achieve sustained improvements in operational effectiveness and customer engagement. We would be delighted to partner with you to improve performance. To start a conversation, please email us at hello@atlanticinsight.com, hit the Contact button on this website or call us on +44 (0) 161 438 2009.

More Insight