Obtaining Accurate Facebook Metrics

“Analytics” has not been getting the best reputation this month. It started with poor election forecasting that left the country questioning the reliability of metrics in front of them. Now at the end of the month, Facebook admitted they inaccurately published metrics and reports of time spent on articles, organic reach for business and more.

However, Facebook’s news wasn’t news to us.

Back in July we discovered some of these Facebook reporting inconsistencies when we could not reconcile one of our client’s Facebook generated insight reports with calculations we gathered through our custom built system.

The good news is that marketers don’t have to compromise accurate data-driven social media planning just because their Facebook metrics are inconsistent. We developed a system and structure to gather accurate Facebook data for multiple brand pages.

Tech Specs

One of our Fortune 500 clients has a large and active global social media community and needed help measuring effectiveness and identifying marketing opportunities on Facebook, Google, Twitter, etc. Here’s the technical side of what we did:

• Developed algorithms and data structures with capabilities to dynamically loop through and count the number of comments and replies across numerous Facebook pages.
• Developed architecture for extracting data from various social-media sites.
• Architected queries using developer tools provided by the social media channels and Qlik Web Connectors.
• Automated the process of daily data extraction from social media, ecommerce, ERP systems, and other data sources.
• Built efficient integrated models for data visualization.
• Designed dashboards with trends and key performance metrics for end-users.
• Researched and resolved the nuances with how APIs are developed and how social media sites change their definitions over time (for example Organic and Viral Impressions in Facebook).

Like many marketers, insight into channel effectiveness drives where dollars get spent. Our systems allowed the client to gain insight into their campaign effectiveness so they could improve their marketing activities to increase their audience’s engagement.

Encountering The Problem

With our system in place, it was time for validation. Since the marketers are using these insights to make budgetary and campaign decisions, this is an important step to ensure the metrics we’re providing are reliable.

One of the specific metrics we worked on was the amount of comments and replies for their postings across many of their brands’ Facebook pages.  We thought we could validate these results with the Facebook provided insight reports (see image below) and the counts shown on each Facebook post.  But this is where the dilemma occurred.

Both the Facebook generated insights reports at the brand level and the counts shown on the Facebook post level did not match our computation. Initially our client assumed that Facebook Insight metrics was accurate and that we needed to review our work.

The Accurate System Is…

We put in extra effort to validate our algorithms and then manually counted the comments and replies at the Facebook post level. This confirmed the accuracy of our system and proved to the client that the Facebook insight report numbers were wrong. This built a great trust and deepened our partnership with our client because we took the time to truly understand their community and needs.

Since we knew exactly what our client needed, we ensured that the metrics were computed in a way that was both accurate and meaningful for their decision making. The benefit with our approach is that our dashboards allow a client to view consolidated results from multiple Facebook pages for different brands and draw comparisons across them.

By developing the algorithms to dynamically extract, integrate, and compute social media metrics across many brands, regions, and sites, our client can access data, find insights and make strategic marketing decisions within their dashboard in a matter of minutes that would have taken weeks to manually extract, verify, and summarize.

By working closely with one client at a time we can obtain rich understanding of their community’s experiences and can better relate to and more accurately develop computations.