Shifting budget toward incrementally effective PPC areas for WWF
WWF have historically implemented a strategy of ‘double bidding’ on keywords within PPC and Affiliate on Pure Brand terms. While this strategy captures all Search Pure Brand activity, the team at Merkle wanted to understand if it was efficient, incrementally impactful on KPIs, and whether there were any cost-savings to be made. With WWF being a nationally recognised company, the aim was to test the value of Pure Brand supported by data.
As WWF operate in the UK, the team proposed a geo-split test comprised of four test regions. The aim was to test the singular and combined effects of Affiliate and Merkle activity on KPIs and to see if budgets could be shifted into the best performing test group. We used Generalised Marketing Areas (GMAs) to split the UK into mutually exclusive groups as GMAs consider commuting journeys between regions and cities, which minimises test group contamination.
The four test groups were:
- Both Merkle and Affiliate bidding on Pure Brand
- Only Merkle bidding on Pure Brand
- Only Affiliate bidding on Pure Brand
- Neither Merkle or Affiliate bidding on Pure Brand
Comparison of conversions across test regions were undertaken weekly during a test period of four weeks until statistical significance was reached – in other words, when a significant difference between the groups was found. Off peak testing was necessary to ensure seasonality wasn't skewing results, and that test insights were obtained prior to the client’s peak period.
Group 2 'Only Merkle bidding on Pure Brand' was the only group that was statistically higher than the other regions. By re-investing Affiliate spend towards the best performing test group, we achieved a 44% increase in Return on Ad Spend (ROAS) without any negative impact on conversion volume. Traffic driven by pure brand increased by 32% and CPC saw a reduction of 51% following from a 36% cut in spend. The efficiency gains were clear, meaning that the most efficient spend for Pure Brand was found through data.