A naive general forecasting model to help set performance targets for websites

As part of my role at FreshEgg I get asked to look at various data driven aspects of the business. Recently a team asked me how they might use past performance data from a client website to help set internal targets for performance for the coming year.

Whilst the client was very happy with the site performance (traffic levels, conversions, etc) the agency wanted to strive for continuous improvement. To help focus the team we wanted to explore how we might set some targets that were realistic and achievable based on what we know about the site’s historical performance and seasonality.

I struggled at first because this client had a business that was not only seasonal but which was very market driven. The client works in the UK residential property sector and as such their website performance is driven by conditions in the wider housing market. As such this makes predicting future performance difficult.

However, interest rates have been relatively stable the last few years, and a look at some top line website statistics year-on-year showed that for the last 4 years traffic have followed a very similar pattern or shape over the year, but the absolute levels increasing slightly each year.

As such for some metrics (such as number of sessions) a year-on-year percentage increase seemed like a sensible way forward, but what would eb a realistic level.

Well to try to help I built a naive forecasting dashboard that allows you to select any metric (such as sessions, conversions, etc) and it will show you year-to-date year-on-year performance. It will then let you plug in a percentage target and it will show you