31 March 2017
Guest blog from Richard Bussy, UK&I Measurement Lead at Facebook
Since advertisers began the quest for marketing effectiveness, the most frequent question asked of publishers and researchers has been what an effective campaign looks like. With dunnhumby Sales Impact, we are now able to measure how effective online activity can be at driving in-store and online sales.
Since we started working with dunnhumby, our aim has always been to help advertisers understand the effectiveness of their online FMCG campaigns in both online and offline sales. dunnhumby have now analysed 37 Facebook and Placement Optimisation campaigns, allowing us to investigate this question through meta-analysis for the first time. This analysis provides cutting edge insight and actionable data to help advertisers optimise their media strategies to drive greater sales impact from Facebook campaigns.
Facebook exposure drives sales
Across these 37 campaigns, dunnhumby saw an average brand sales lift of 3.7%1, which resulted in an average lift of £14.93 per 1,000 households2. In the analysis there are two sales levers that dunnhumby looks at:
1. the household penetration of the product or brand.
2. the spend towards the product or brand from buying households.
dunnhumby found that both household penetration and buying household spend increased following exposure to a campaign3, with penetration increasing on average by 3.8%, and buying household spend increasing 0.9%.
To give actionable insight to marketers and planners, dunnhumby have identified three key factors in driving sales impact; reach, frequency and duration.
Last year we published a paper on reach, based on US CPG campaigns4, showing that campaigns in the top quartile of reach drove penetration and incremental sales. dunnhumby was able to look at their data in a similar way, revealing parallels in the UK market, albeit with a slight difference.
Although dunnhumby didn’t have the quantity of campaigns to split by quartiles, they were able to split the campaigns into halves, of reach above 8 million people, and below 8 million people. Although this isn’t a perfect proxy for broad targeting and niche targeting, in many of the campaigns this was how the split worked, and as such the findings reflected the powers of targeting.
Campaigns that were focused on less than average reach (which in many cases was a focussed audience) produced a slightly higher household penetration, but broader-than-average reach campaigns generated a slightly higher household spend (2.6X) and a far higher return on ad spend (9.4X), due to the fact they targeted a far greater number of people. This doesn’t mean that the same creative will work with every audience, but by reaching as large a relevant audience as possible (and using relevant creative) produced the greatest return in this analysis.
Effective frequencies increase impact
A paper published last year was able to demonstrate that the effective frequency of a campaign is dictated by its business objective. It found that, typically, campaigns seeking ad recall could be effective with a lower frequency, while campaigns seeking purchase intent needed a higher frequency. However, there is no ‘correct’ frequency level; a variety of market and environmental factors are present, and where an advertiser’s campaign sits within these will lead to suggestions for the frequency level being dialled up or down to achieve success.
dunnhumby looked at their data to understand the effect of frequency on campaign performance. Grouping campaigns by frequency levels5, they found that higher frequency campaigns drove greater household penetration, greater spend from buying households and ultimately drove a greater sales lift overall for the product or brand.
Furthermore, when combining frequency and duration, split as less than 6 weeks and 6 weeks or more, they saw that the campaigns studied that were allowed to build their frequency at an effective level over a longer period achieved the greatest level of sales lift.
What does this mean for advertisers?
dunnhumby’s analysis shows that advertisers who planned campaigns with a relevant but broad reach, running for more than 6 weeks, at a frequency above 1 per person per week saw the greatest benefit from including Facebook on their media plans.
Combining these findings with those of the papers published last year, and applying the adjustment factors referenced, we are starting to see that for a campaign to achieve optimal sales lift it should ideally be planned with a frequency ≥1 per person per week (not to be confused with f=1), across a duration of more than 6 weeks.
If an advertiser knows that people outside a core audience will not buy their products – perhaps due to them only being relevant to one life-stage or one gender – then narrow targeting could be optimal, but more generally the rule is that a far greater return on ad spend will be achieved by aiming for as broad a relevant audience as possible.
To learn more about linking online campaigns to offline sales, click on the button below to download our Sales Impact meta-analysis report
 dunnhumby Sales Impact is built with an unexposed vs. exposed methodology. Sales lift is the percentage uplift from the unexposed group to the exposed group.
 For comparability, wherever possible we look at uplift in percentage rather than value terms due to the differences seen between categories. For instance, a 5% uplift in the food category will give a lower value uplift than a 5% uplift in household care, as the typical product costs less money. For this reason, dunnhumby focuses on percentage uplift wherever possible, unless the value is needed to tell an important part of the story.
 dunnhumby measures campaigns run on Facebook, Instagram and via Placement Optimisation (across both Facebook and Instagram)
 Consumer Packaged Goods, similar to FMCG, or Fast Moving Consumer Goods, in the UK
 “Lower frequency”, or “f<1 pppw”, and “Higher frequency”, or “f≥1 pppw”