by Carly Sharma, Marketing and Data Science Analyst, LiftLab
November 2, 2023
Recently, we hosted our 2023 Customer Summit, where we had the opportunity to engage with our customers on several topics—one of the standout moments focused on the important topic of Influencer Marketing and how to measure its impact. Every marketing organization we spoke with was trying to assess this.
In Influencer Marketing, the potential for high engagement and brand exposure is undeniable, but the path to assessing the actual impact of influencers remains tangled and often nebulous. All our participants were convinced influencer marketing was valuable but agreed that measuring the incremental benefit from these channels was one of their team’s more difficult challenges. Our Customer Summit illuminated the challenges and strategies in this domain, but the path to perfection remains iterative.
Ideal data for the LiftLab Platform
We started our discussion with the end goal in mind.
“How do you pinpoint the perfect data set that LiftLab would want to analyze your influencer marketing outcomes effectively?”
While the initial question seems straightforward, the larger issue is whether your team can provide this data.
So, what is the Ideal Data Set?
Why measuring influencer data is so difficult
Here are some of the insights from our community:
There are a lot of blurred lines. How do you demarcate between organic vs. paid social vs. affiliate programs? And what if you amplify a paid influencer post by turning it into an ad? Some teams struggled to define influencer marketing vs. other paid social and affiliate channels.
Does this fall within a more creative team or an analytical team? If this falls on more of an operational/creative team, how do you bring an analytical structure to the team? Some clients came from teams that started with the execution side of influencer initiatives and worried about measurement later.
How do we determine how long we expect the drag effect to last for each channel/post?
If you are tracking each influencer, some channels have lower engagement with those links, but could still be very important. For example, vanity links for podcasts.
Current strategies our community uses that, while imperfect, start to get you to an answer:
Calculating daily spend: Typically, influencers are paid in product or up front, not daily. For a 24-hour Instagram story post, this is not a problem, but for channels with high drag, like YouTube, you need a way of calculating the daily spend. Drag is the lingering effect of a marketing activity, even after the activity has ended. It helps calculate how the audience is reached rather than just the cost of the ad. If your team has determined that the post has its maximum influence for seven days, divide the spend by seven days.
Tracking engagement in multiple ways: If you can use a tag in or UTM, that is a great start. Some of our clients also used other methods, like surveying customers when purchasing. Or, having a team that signs into a channel, like YouTube, every day to see how many new impressions a post receives. This way, you have daily impressions and daily spend to run a model. However, engagement doesn’t equate to conversions. Defining success for your business, whether brand awareness or purchases, is essential to estimate whether the ROI is worth it.
Know your CPA to measure success against: Even if measurement is challenging, our customers with the most success ultimately answer the question, “If we pay an influencer $XXXX, how many conversions did she/he drive? Did that beat our CPA for that channel?”
Which kind of influencer to choose? There are different tiers of influencers, and some may be right for some businesses vs. others. To simplify, an extremely expensive but very influential influencer, like a Kardashian, may drive more conversions because their followers really listen, but they are costly. On the other end of the spectrum are micro-influencers who are affordable. Sometimes, they are less curated regarding which brands they will work with, watering down the influence each post has. Then, there are many in between. It’s essential to consider these when selecting influencers.
Where LiftLab fits in: There are a couple of options using the LiftLab Platform when it comes to experimenting with influencer data.
If you run ads featuring influencers, you can run a common LiftLab “Go Dark with Pacing” experiment. This type of experiment involves grouping similar market segments into test and control groups and varying spend in the test group. During a specified timeframe, ads are displayed in these test markets. Because paid social allows you to filter by location, this experiment type is available.
When dealing with organic social channels, like influencers sharing content on their own accounts, you lack control over the locations where the content is seen. Such posts can be seen by anyone, anywhere, regardless of location. Our clients have had success running “Switchback” experiments in these instances. An effective strategy involves having multiple influencers post content on the same platform within a short time frame at the start of the experiment window, then stop, then post again a couple of weeks later, repeating until the experiment ends. This approach allows you to measure the cumulative impact of influencer posts on that particular channel. It’s important to remember the channel you run the switchback test on to guarantee a distinct “off” period. For example, the temporary nature of Instagram’s 24-hour posts allows an influencer to post up until the day before an off period. However, for TikTok or YouTube, if you assume there is a long drag period of around seven days, you’d want the influencers to all post at the beginning of the two-week-on period only to capture the drag effect and not have it spill over to the off period.
As we keep the dialogue going with LiftLab clients, we will continue to find new recommendations and share them with our community. Looking forward to the dialogue!
Carly Sharma began her career as one of the first five employees at Airbnb and subsequently ran her own profitable startup for seven years. Always interested in data, she more recently became a data scientist, and she now helps LiftLab with various R&D projects and experimentation.