Table of Contents
Measuring Social Media ROI in 2023
Prominent figures like Les Binet emphasize the need for a balanced approach between long-term brand building and short-term activation. While conversions are tempting to track, the constraints of attribution modeling make this an unreliable metric on its own. The latest techniques combine brand lift studies, incrementality testing, and multivariate analysis to uncover meaningful correlations that approximate the true value of social.
Causal Impact Analysis
Causal impact analysis is growing in popularity for social media measurement. This methodology was originally developed by Google and calculates the effect of an intervention on a time series by comparing the actuals against a synthetic control. For social media, the intervention could be launching a new campaign or increasing posting frequency. The time series is a business KPI like website traffic, leads, or sales. Causal impact determines whether lift occurred by testing the statistical likelihood that the difference between the synthetic control and actuals was due to random chance after the intervention occurred.
Some benefits of causal impact analysis include:
- Accounts for seasonality and natural fluctuations in the time series, making it more statistically robust than simple before-and-after comparisons.
- The synthetic control provides a counterfactual estimate of what would have happened without the intervention.
- Useful for ongoing optimization as new interventions can be continually tested against the evolving synthetic control.
- Quantifies the incremental lift generated, including a confidence interval.
To implement, historical performance data is needed over a sufficient time period to train the synthetic control. Many data scientists recommend at least 50 time periods if measuring a weekly time series. Marketing scientists have applied causal impact analysis to quantify the true performance of social campaigns on metrics like website traffic, coupon redemptions and call volume with positive results.
Incrementality testing is another methodology growing in adoption to measure channel attribution. This approach aims to determine the incremental impact of a marketing channel by running an A/B test with the channel turned on for the treatment group and off for the control group. By randomly assigning users to each group and keeping all other conditions equal, any difference in behavior can be attributed to the impact of the channel being tested.
For social media, running held-out tests can uncover performance on metrics like registrations, purchases and survey participation.
Some best practices for incrementality testing include:
- Holding out a statistically significant portion of spending, such as 10%, rather than pausing entirely to minimize business impact.
- Running tests across multiple time periods and user segments to validate results.
- Comparing performance over a sufficient time period to account for lag effects.
- Applying learnings to optimize social programs for greater efficiency moving forward.
Social media plays an important brand building role that is difficult to capture through direct response metrics alone. Robust methodologies like brand lift studies can uncover how social activity changes consumer perceptions, attitudes and affinity.
Some examples of what can be measured through multi-cell brand lift studies include:
- Brand awareness: Exposed vs control group aided/unaided recall
- Brand favorability: Exposed vs control group ratings on brand attributes
- Message association: Exposed vs control group attribution of slogans or campaigns
- Purchase intent: Exposed vs control group self-reported likelihood to buy
Digital providers have made brand lift studies more accessible through survey sampling and measurement on social media platforms themselves. By determining social media’s ability to drive brand lift on key perceptions, marketers can better quantify its value.
Multivariate testing uses statistical modeling to uncover correlations between multiple independent variables and a business outcome. For social media, running controlled experiments across different content types, formats, calls-to-action and targeting parameters can shed light on what posts, audiences and strategies drive the most impact.
Some benefits of multivariate testing for social media include:
- Determines interaction effects between different variables that impact performance.
- Models can determine optimal social contentmix for different audiences.
- Provides learning agility to continuously adapt and improve content strategy.
- More statistically rigorous than changing one variable at a time.
The MODEL framework is a helpful approach for designing multivariate tests:
M – Metrics: Identify key business outcomes to measure
O – Objectives: Hypothesize which variables will influence outcomes
D – Design: Create test posts with different variable combinations
E – Execute: Publish test posts and measure outcomes
L – Learn: Analyze results to identify most impactful variables
Measuring social media ROI requires an evolution from simple conversion tracking to methodologies that can approximate causal relationships and incremental value. While no approach provides perfect accuracy, combining experimental frameworks like incrementality testing and causal impact analysis with brand lift studies and multivariate testing enables a more robust view of social’s financial impact. Rather than trying to attribute every downstream sale, marketers should focus on high-quality analytics that connect social activities to tangible business results. With privacy regulations limiting traditional tracking, statistically-based methodologies represent the future of social media measurement and can help maintain budget for an influential channel.