Reach Matters: Driving Business Results at Scale - Facebook IQ

White paper

Reach Matters: Driving Business Results at Scale June 2016

Contents

1 2 3 4 9

11

Executive Summary Introduction Why reach matters Part I - Can scaling reach drive additional sales? Part II - What’s the best approach to driving higher reach? Key findings – Distribution of share of people reached – Distribution of brand impact – Cost efficiency of driving brand impact

15

Conclusion – Takeaways

Executive Summary

Digital and mobile platforms, like Facebook, have given marketers the opportunity to drive both precision to target audiences as well as mass scale to reach larger ones. With the power of precision, are marketers losing out on the opportunity to drive impact at scale?

What it mean for marketers To explore the importance of utilizing reach on digital, the Facebook Marketing Science team conducted research to determine how efficient reach is in driving impact and real business results at scale for brand advertisers. Our findings show how scaling reach can cost-efficiently impact more people, change their attitudes and behaviors and increase in-store sales. Reach and real business results: Facebook worked with Consumer Packaged Goods companies over the past three years to understand the impact of their Facebook media spend on in-store sales. Through a partnership with Oracle Data Cloud we were able to anonymously match users who saw a Facebook or Instagram ad to their in-store purchases for 178 campaigns over 35 different CPG advertisers. For part of this analysis, we compared the top reach driving quartile to the bottom reach driving quartile of campaigns. We saw that the top reach quartile of campaigns drove 4(2$$4)-$3 the number of total people impacted at 10% less cost per person impacted. The top reach quartile also drove 139% more incremental sales and 169% more incremental buyers.

Reach and brand impact: How can marketers implement a broader reach strategy in their campaign execution? On Facebook and Instagram, campaign execution is based on an auction where advertisers optimize their campaigns based on different objectives. In our study, we conducted a meta-analysis of 57 large brand advertisers and found that marketers who employ bidding strategies focused on driving reach can more cost-efficiently drive brand impact (Ad Recall and awareness) at scale and impact more people. Read on for more details on the study, its design and our conclusions.

1

Introduction

Marketers have the ability to control who they want to reach and how they spend their media dollars. But with the power of precision has come the trade-off of reach. An advertiser’s audience is often dictated by the marketing medium itself, especially when it comes to traditional media. Take, for example, primetime TV, which reaches very broad, general audiences. Or billboards, which reach audiences within a geography that is likely broader than the brand’s desired target. With digital and mobile platforms, like Facebook and Instagram, marketers have the ability to narrow in on target audiences and drive mass reach to larger ones. By narrowing in on target audiences, are marketers losing out on the opportunity to drive impact at scale?

2

In this paper, we’ll look more closely at a common situation marketers are facing today: deciding how to balance precision and scale and how these decisions affect total impact. We’ll then explore research that uncovers the opportunity marketers have when scaling reach to drive additional sales. Finally, we’ll analyze how marketers can achieve greater reach on Facebook and Instagram in terms of campaign optimization. Through this analysis, we’ll help advertisers understand how scaling reach increases their opportunity for incremental sales, decreases their cost per person reached, and drives the most incremental sales and attitudinal impact.

Facebook IQ White paper

Why reach matters

Advertisers have to balance between relevancy and scale. For most brand advertisers, is optimizing towards reach the most effective lever they can pull?

In broader reaching campaigns, even small lifts can result in impacting a greater number of people.

The digital space offers advertisers the ability to define their target audience using a variety of signals, such as demographics, interests, or behaviors. Doing so may result in a higher level of impact to purchase or change consumer opinions about the brand. But this may also lead to a trade-off between the overall scale of the campaign and the potential number of people impacted. Advertisers could be limiting their reach based on the targeting criteria, scale of budget, or the method for purchasing and optimizing media

(i.e.ř the impact of advertising) is relatively small. In the second scenario, the campaign is targeted more broadly and may reach some of the same audience as the first example, as well as additional people. Even though we see a lower response rate to the advertising, the total number of people impacted is higher. For most advertisers, the key objective should be measured by the total number of people impacted. In this case, assuming equal budget, the broader audience execution would be more successful.

To illustrate the balance advertisers face between relevancy and scale, we looked at two scenarios. In the first scenario, the advertiser is reaching a smaller and potentially more relevant audience. While the response rate of the advertising is higher, the total number of people influenced by the campaign

These examples illustrate the balance advertisers face between precision and scale when planning their campaigns. While very precise targeting may lead to higher response within that audience, understanding these tradeŬoffs advertisers face is an important consideration.

Consider the following illustrative example of two scenarios for a campaign execution

Target audience

Lift

Precise, smaller audience

12%

Broader audience

8%

Audience impacted

3

PART I - Can scaling reach drive additional sales?

Keeping in mind the total number of people impacted, we wanted to understand if these scenarios hold true with realworld campaigns. That is, if we look at groups of campaigns with different reach levels, how can we translate total impact of a campaign to understand the differences in number impacted, the cost of the media per person impacted, and the overall opportunity marketers have to scale reach.

4

Methodology Over the past three years, Facebook has been working closely with Consumer Package# Goods (CPG) advertisers in the US to help them understand the impact their Facebook media has on their brand’s in-store sales. Through a partnership with Oracle Data CloudŉŊ, advertisers can match anonymized groups of users who saw an ad on Facebook or Instagram to their subsequent in-store purchases at partner retailers. Using hashed data, this enables Oracle Data Cloud to measure the causal impact of advertising campaigns run on Facebook and Instagram.1 Facebook recently partnered with ODC to analyze Facebook campaigns measured between January 2014 and June 2015. This meta-analysis focused on the results of 178 campaigns that were measured in this time period, across over 35 different CPG advertisers. These campaigns were used as the base to understand what factors helped drive in-store sales performance.


Key findings

Can scaling reach online drive incremental brand sales offline?

Reach drives in-store sales for CPG2 Top quartile campaigns in driving incremental sales had almost twice the average reach than the bottom quartile of campaigns

22mm reach

12mm reach

Bottom quartile sales

Top quartile sales

First, we wanted to understand how the best- and worst-performing campaigns drove incremental brand sales. Oracle Data Cloud ranked the campaigns based on incremental sales and compared the top quartile (the top 25% of campaigns) to the bottom quartile (the bottom 25%). The study found that the top quartile of campaigns had almost twice the average reach compared to the campaigns in the bottom quartile. Campaigns in the top quartile drove 75% of the total incremental sales observed across all 178 of the campaigns and displayed higher increases in incremental trial purchases and household penetration.

5

Key findings

Top quartile of sales driving campaigns3 Top quartile of campaigns drove 75% of the sales observed across all the campaigns. Top quartile

28x

10x

the lift in sales

2x

the lift in penetration

8x

the lift in new buyers

the reach

Bottom quartile

In addition to understanding the differences between the top and bottom reach-driving campaigns, Facebook also looked to understand the effect on the total number of people impacted. Of the 178 campaigns outlined above, we looked at 122 examples where advertisers provided media cost data to Oracle Data Cloud and that had statistically significant sales lift. We then separated these 122 campaigns into quartiles by reach.

6


Key findings

Focusing on the median results of the top and bottom quartilesŘ we observed the following: Median reach

Median point difference in buyers (penetration)

Median campaign cost

Cost per reach

Bottom Reach Quartile

8 million

0.09%

100 (index)

100 (index)

Top Reach Quartile

32 million

0.06%

240

60

As reach is increased, we see that campaign spend increases. We also see that observed point difference lift in penetration, or buyers, is lower in the top reach quartile. Cost per reach is also much lower in the top reach quartile, as advertisers who may have optimized towards higher reach had opportunity to reach

cheaper people on Facebook and Instagram. This is discussed further in Section III of the paper. Taking this information, we calculated the total number impacted, or in this case, converted, from these scenarios.

Median reach

x Median point difference in buyers (penetration)

= Total people impacted

Bottom reach quartile

8 million

0.09%

7,520

Top reach quartile

32 million

0.06%

20,240

We see that for the top quartile of campaigns, reach is higher, while point difference in buyers is lower. This translates to more than twice the number of people impacted. For marketers who had high reach

campaigns, they were able to drive change in buying behavior from a larger number of people. That said, how well does the high reach strategy pay off?

Median campaign cost

/ Total people impacted

= Cost per impact

Bottom reach quartile

100 (index)

7,520

100 (index)

Top reach quartile

240

20,240

90

Finally, we see that the cost per person impacted is lower in the top quartile of reach than it is in the bottom quartile, making it the more efficient strategy. Looking at these top and bottom quartiles, we observed that the median incremental sales of the top quartile was 139% higher than the bottom quartile. The lower reach quartile may have had a more precise

audience, which has more response per person, but the top reach quartile drove more people to buy. This shows that, for marketers, focusing on reach drives more buyers and ultimately more incremental sales.

7

Key findings

Top reach quartile

169%

139%

More incremental buyers

Higher campaign incremental sales

The findings of this meta-analysis reinforce that optimizing towards reach will help advertisers drive additional campaign performance in terms of incremental in-store sales. Facebook campaigns with the highest sales performance saw significantly more reach on average. Even though the price of a campaign increases along with reach, advertisers have the

8

opportunity to drive more people in-store to buy. This maintains positive return on their media investment while leading to additional incremental sales.


The highest reaching campaigns saw significantly more impact on average.

PART II - What’s the best approach to driving higher reach?

Our research has shown that marketers have the opportunity to increase reach and drive a greater number of people to change their behavior— and to do so efficiently. The next logical question is how a marketer can actually implement a broad reach strategy in their campaign executions. Marketers need to align their campaign execution against their marketing objective. With brand marketers focused on reach, there are multiple levers a marketer can pull to drive greater reach: (1) broaden audience definition, (2)scale media budget or (3) execute their campaigns to optimize and align to their reach objectives. In the case of Facebook and Instagram, where campaign execution is based on an auction, marketers should employ bidding strategies focused on driving reach in the most efficient way. Facebook’s auctionbased mediaŬbuying environment is executed real-time, with an auction taking place whenever someone is eligible to see an ad.4 The participants in an auction represent competing ads targeted to the audience the eligible person falls into, and the winning participant is the ad that creates the most overall value. Billions of these auctions take place every day. In this section, we will explore how marketers can optimize the execution of their campaigns to increase overall reach and reduce cost per reach. We will then look at this in terms of the total people impacted to understand how bidding and buying strategies can drive greater results for marketers.

Reach Matters

Methodology To dive into the question of execution, we will focus on attitudinal brand metrics as the indicator of success. Similar to sales metrics, we can look at brand metrics in terms of incremental people impacted and the cost efficiency of driving this impact. Within Facebook’s auction-based media-buying environment, marketers’ bidding strategy decisions impact overall campaign reach, the types of people reached and the cost efficiency of driving results. In our analysis, we estimated the impact of bidding strategy by understanding who we are reaching and how much it costs to reach them. We focused on two broad bidding strategiesŚ reachoptimized bidding and action-optimized bidding: Reach-optimized bidding: comprised of campaigns bidding for daily reach, cost per thousand impressions (CPM) and Reach and Frequency.5 Action-optimized bidding: comprised of campaigns bidding for cost per click (CPC), including post engagement, website clicks and website conversions.

9

PART II - What’s the best approach to driving higher reach?

To understand the relative cost of reaching people across the two bidding strategies, we first conducted a meta-analysis of 57 large US brand advertisers’6 Facebook campaigns that were in-market from October to November 2015, and calculated userlevel CPMs for people reached by these campaigns. We calculated the share of people reached by CPM cost ranges (CPM quintiles7) for campaigns with reachŬ and action-optimized bidding. People reached by the advertising campaigns were categorized into five groups or quintiles, depending on how expensive they were to reach on Facebook, with Quintile 1 representing the group with the lowest user CPMs and Quintile 5 representing the group with the highest user CPMs.8 This information was then combined with the observed brand impact on the five groups reached based on a meta-analysis of brand survey experiments9 conducted on Facebook and Instagram.

10

For the second meta-analysis of brand survey experiments, conducted between September 2015 through November 2015, we analyzed brand survey data associated with US Facebook and Instagram brand experiments from a cross-section of verticals. We analyzed results for brand experiments (lift) by user CPM quintiles to understand the distribution of brand lifts across different groups of people based on their CPMs. We focused on understanding the distributions of two types of brand lifts: Ad Recall (437 experiments) and brand or campaign awareness (213 experiments). By combining findings on the share of people reached and brand impact results by CPM quintiles, we can understand the overall brand impact and the cost efficiency of driving that impact between reachŬ and action-optimized bidding.


There are multiple levers a marketer can pull to drive greater reach: (1) broaden audience definition, (2) scale media budget or (3) execute their campaigns to optimize and align to their reach objectives.

Key findings

Distribution of share of people reached

as we will discuss later in this paper, reaching the most expensive people doesn’t necessarily translate into the most cost-effective strategy for brand advertisers looking to drive brand impact at scale.

In our meta-analysis of Facebook campaigns from 57 large brand advertisers, we found that the average click-through rate for people in the most expensive CPM quintile was 4X to 5X that of the least expensive CPM quintile, regardless of bidding strategy. However,

This pattern suggests that reach-optimized bidding allows marketers to reach a greater share of people with lessŬexpensive user CPMs. In other words, reachoptimized bidding allows marketers to have a lower cost per reach compared to action-optimized bidding.

The chart below compares the distributions of the share of people reached by the campaigns in our study with reach-optimized bidding and action-optimized bidding. The distribution for reach-optimized bidding is steeper from left to right compared to action-optimized bidding. Given these differences, we found that the average user CPM for action-optimized campaigns was 80% higher than for reach-optimized campaigns.

Figure 1: Comparison of share of people reached by campaigns based on bidding strategies

40%

30% Share of reached (%)

Action-optimized campaigns resulted in an 80% higher CPM than reach-optimized campaigns, on average

In Facebook’s auction-based media-buying environment, marketplace dynamics—demand and supply factors, among others—determine the relative cost of reaching people within an advertiser’s target audience at any given time. A variety of factors can influence whether people are more or less expensive for an advertiser to reach. Ongoing internal research has shown that one important factor is “user clickiness”—the relative propensity of people to click on ads. We find that clickier people are more expensive to reach compared to less clicky people because of greater competition from advertisers trying to reach the same group of people, among other reasons.

20%

10%

0% 1

2 Reach

3 CPM quintiles

4

5 Action

Source: Share of reached people on Facebook by user CPM quintiles based on Facebook mobile feed impressions for 57 large US brand advertisers for campaigns that were in-market during October to November 2015.

11

Key findings

Distribution of brand impact

We’ve seen that the mix of people reached can vary with reach- or action-optimized bidding. That said, are these groups more or less likely to be impacted by brand messages? To answer this question, we analyzed brand survey data from Facebook and Instagram brand experiments and focused on understanding the distributions for 2 types of brand lifts: Ad Recall and !rand or "ampaign wareness. The following charts show the distributions of lift in Ad Recall and wareness by user CPM quintiles. The brand lift results are indexed to the average lift in the dataset for each category. We see that Ad Recall lift is positively correlated with user CPM quintiles—a finding that is consistent with previous internal

research. On the other hand, lift in awareness is relatively flat in relation to user CPM quintiles. These results suggest that people who are more expensive to reach are likely to better remember seeing ads on our platform. One hypothesis is that these people are also clickier and therefore likely to engage more with the ads. However, the results for awareness implies that people who are more expensive to reach (and presumably clickier) aren’t likely to respond better in terms of awareness of the brand or campaign. That said, what is the overall impact of brand messages by bidding strategy? We can combine our findings on

Figure 2: Ad Recall lift by CPM quintiles

Ad Recall lift (indexed)

150

100

50

0 1

2

3

4

5

CPM quintiles Source: 437 brand experiments measuring ad recall conducted between September and November 2015 in the US. Experiments with at least 200 responses in the test and the control groups were used in the analysis.

12


Key findings

Figure 3: Awareness lift by CPM quintiles

Awareness lift (indexed)

150

100

50

0 1

2

3

4

5

CPM quintiles Source: 213 brand experiments measuring brand or campaign awareness conducted between September and November 2015 in the US. Experiments with at least 200 responses in the test and the control groups were used in the analysis.

the share of people reached and brand impact results by user CPM quintiles to estimate the average lift in Ad Recall and awareness for reach- and action-optimized bidding. We estimated that on average: •

Ad Recall lift for reach-optimized bidding is 0.9X that of action-optimized bidding

•

Brand or campaign awareness lift for reachoptimized bidding is similar to that of actionoptimized bidding

results. Although the mix of people reached varies by bidding strategy, the overall impact of brand messages is fairly similar.

We see that for both brand resonance metrics, reachand action-optimized bidding have similar overall

13

Key findings

Cost efficiency of driving brand impact

We saw that the overall impact of brand messages in terms of Ad Recall and brand awareness are fairly similar for reach- and action-optimized bidding. However, when we consider the cost per person impacted, reach- and action-optimized bidding can produce different results. Let’s assume that a brand advertiser plans to run an eight-week campaign with $500,000, with average campaign CPM at $4 for reach-optimized bidding, and campaign frequency at 3 impressions per person per week. Because we found that the average lift in awareness is similar across reach- and actionoptimized bidding, let’s assume that the expected brand awareness lift is 4% for both strategies.

A simulated Facebook campaign based on our findings Reachoptimized

Actionoptimized

Campaign spend

$500,000

$500,000

Average CPM

$4.00

$7.20

Reach

5,200,000

2,900,000

Expected lIft

4%

4%

Total impact

208,000

116,000

Cost per impact

$2.40

$4.31

Assuming that the average CPM for action-optimized bidding is 80% higher than reach-optimized bidding,10 the brand advertiser gets a larger reach (5.2 million people) with reach-optimized bidding compared to action-optimized bidding (2.9 million people). Therefore, given that both optimization scenarios have the same campaign budget, the same frequency and the same expected lift in brand awareness, reachoptimized bidding drives more scale and is therefore more cost-effective in terms of generating brand awareness. Specifically, the cost of driving incremental change in brand awareness is $2.40 per person with reach-optimized bidding versus $4.31 per person with action-optimized bidding. By optimizing towards reach, the advertiser can more cost efficiently drive awareness for the brand.

14


Based on this simulated campaign, reach-optimized bidding was 1.8x more cost efficient and drove greater total impact than the action-optimized campaigns.

Conclusion

“Reach is of critical importance when planning campaigns that drive our business. By planning and optimizing campaigns towards reach on Facebook, we’ve been able to see continued success in driving in-store sales.” Atin Kulkarni, Senior Director, Portfolio Analytics, Frito-Lay North America With digital and mobile channels giving advertisers the opportunity to fine-tune who they want to reach, media buying may result in a strategy too limited to specific, narrow groups of people. This may not be the best approach to maximize campaign impact and drive a marketer’s business. Our research demonstrates that reach has always— and will continue to—be an important driver of campaign effectiveness. Our meta-analysis on digital campaigns and in-store sales in partnership with ODC shows that optimizing towards reach helps advertisers drive positive ROI performance at scale. In terms of campaign execution on Facebook and Instagram, our meta-analysis on brand resonance experiments shows that reach-optimized bidding allows advertisers to drive brand outcomes more cost efficiently compared to action-optimized bidding. By focusing on scaled

reach—whether by increasing audience sizes, increasing budgets or executing campaigns aligned with reach objectives—brand marketers can drive higher amounts of incremental sales and attitudinal impact more efficiently.

What it mean for marketers As marketers have more options and more control over how their media budgets are spent and optimized, especially on digital, sacrificing reach for precision may not always be the best strategy. For brand marketing, changing attitudes and behaviors at scale will ultimately lead to growing more business. Considering the implications of targeting, budget allocation, and execution optimization when planning towards reach objectives will allow marketers to impact the most people possible in the most efficient way.

TAKEAWAYS: 1.

Plan to your objectives. For brand marketers, having reach as a key objective to guide execution will drive business at scale.

2.

Measure campaigns in terms of total impact which translates to number of people influenced (change in behavior or attitude) and cost per impacted person. Lift metrics alone may be misleading because they don’t factor in the scale of the campaign.

3.

Reaching more people can be an overall more efficient strategy for brand marketers. For an auction setting in particular, optimizing media buying practices and execution tactics towards reach versus clicks or actions will ensure the lowest cost per reach and cost per person impacted.

15

Appendix

Notes & Sources

More on how Oracle Data Cloud helps advertisers measure in-store sales: https://www.datalogix.com/ measurement-and-insights/. 1

Meta-analysis of 178 Oracle Data Cloud-measured Facebook and Instagram campaigns, Jan 2014– Jun 2015. 2

Meta-analysis of 178 Oracle Data Cloud-measured Facebook and Instagram campaigns, Jan 2014– Jun 2015. 3

For more information, see https://www.facebook. com/business/help/430291176997542 4

Reach and Frequency buying enables advertisers to predictably buy unique reach on Facebook and Instagram, while giving them the tools to control frequency throughout their campaign. 5

These advertisers also ran brand measurement studies on Facebook from September to November 2015. We analyzed a representative sample of impression data for the business-as-usual campaigns of these advertisers to understand their usual behavior on our platform. We also looked at advertisers that had spent more than $100,000 on Facebook over a two-month period as of November 30, 2015. These selection criteria were chosen to ensure that the results for the share of people reached by user-level CPM quintiles could be tied back to brand impact results that are typically measured for the largest brand advertisers.

We calculated the boundaries or thresholds for the user CPM quintiles based on a 1% representative sample of three month active Facebook users in the US as of Nov 30, 2015. 7

Put another way, the lower threshold for Quintile 5 is four times the upper threshold of Quintile 1.

8

In brand survey experiments, people within the targeting set for an advertiser’s campaign are randomly assigned into treatment and control groups. A sample of people in each group are surveyed and their responses compared to each treatment group and to the responses in the control group to determine whether the paid media generated lifts in brand resonance metrics. Lifts are defined as an increase in positive answers to brand survey questions based on a comparison of responses from treatment groups to the responses in control groups. 9

6

16

Here, we’re applying our findings on average user CPMs to overall campaign CPMs.

10
