What does it mean for forecasting, when past data is not predictive of the future?

2 min read
BOSCO™-Blog-5

As we start to approach some sort of normality after the last 12 months, our Principal Data Scientist, Dr James McKeone shares his thoughts on forecasting in 2021.

Looking to the future

The future is bright in 2021. It’s worth considering now what our post-covid media landscape will be and planning for the scenarios that will soon be reality. The end of lockdowns as vaccination rates increase, return to offices, return to international travel and holidays, return to window shopping and physical retail. Answering the “when” of these scenarios seems less important as the “what” we can do once they occur. It will be fascinating to see the extent to which covid has altered behaviour permanently as consumers build have now learned to shop online, work and socialise from home and this will continue to some degree post-covid. Data and model-based media planning and ecommerce strategies are critical to success.

Ask a data science team to forecast sales for the next month-ahead and they’ll likely ask for at least two years data, any less and the team will raise issues about properly capturing year-on-year seasonality. This seems totally reasonable given the seasons, the holiday calendar and many sales and promotions follow this pattern. The issue now given that 2020 was such an outlier in terms of performance is that the forecasting models will need to be built on more short-term seasonal effects.

Fundamentally, businesses must evaluate the extent to which past performance is useful in predicting future performance. Including too much of that history now that consumer habits have changed can result in poorer predictions than might otherwise be the case. The result for media planning is primarily about how we measure success, plan more scenarios, understand the opportunity costs of multi-channel media investment, and be prepared to react quickly when demand changes.

Book a demo

For some firms, the idea of a month-ahead forecast of performance under different budget scenarios can often seem unachievable. BOSCO™ answers the question of where and when to spend your multi-channel paid media budget. Critical to achieving this view is up-to-date telemetry on the recent past and a clear understanding of how changes in one channel can impact others and continually evaluate and review performance for their business. Try BOSCO™ for free or get in touch to book a demo.

Author

Stay in the loop
Loading
Share post

hi

Other posts you might like

How to plan your first-party data strategy: A beginner’s guide

How to plan your first-party data strategy: A beginner’s guide

Privacy regulations and the phasing out of third-party cookies have made it more important than ever to establish a solid
ROI vs ROAS in PPC: What’s the difference and how should you use them?

ROI vs ROAS in PPC: What’s the difference and how should you use them?

TLDR: ROI tells you whether you actually made any money. ROAS tells you how efficient your ad spend was. ROI
Digital news to watch: AI boosts Search Console insights

Digital news to watch: AI boosts Search Console insights

In this month’s digital news, Google Search Central has rolled out a new AI-powered configuration feature within Search Console, aimed

Popular topics

[other_categories]