A pet food brand's social team built a dashboard comparing video performance across five platforms. One platform showed roughly four times the views of the others, so leadership moved a large share of the content budget there.
Two months later, website visits and sales from that channel had barely moved. When an analyst dug in, the explanation was almost embarrassing. That platform counted a view after a far shorter watch time than the others. The videos weren't performing four times better. They were being measured on an easier scale.
Every Platform Counts Differently
Social platforms each define their metrics their own way, and the differences aren't small. A "view" may start the instant a video appears, after a couple of seconds, or only after a longer watch. "Reach" might mean unique accounts or something closer to total impressions. Engagement rate gets calculated against followers on one platform and impressions on another.
Put those numbers side by side in one chart and they look comparable. They aren't. Decisions made on them carry the distortion forward, which is exactly what happened to the content budget.
The Monday Spreadsheet
Before the dashboard, the team built its weekly report by hand. Someone downloaded exports from each platform every Monday morning, pasted them into a master spreadsheet, fixed column names, and built charts. It took about six hours.
Errors crept in constantly. One platform changed its export format without warning, and for three weeks a column of shares was being read as comments. Nobody noticed because the numbers still looked plausible.
The hand-built report also never connected to what happened afterward. Did people who watched the dog-food-taste-test video visit the website? Did any buy? The spreadsheet couldn't say.
Building a Pipeline Instead
The brand's small data team replaced the Monday routine with an automated pipeline. Every night it pulls metrics from each platform, applies consistent definitions, and joins the results with website sessions and orders tracked through campaign links.
They built it in Prophecy, which let them design the transformations visually while the platform generated code into version control. That combination turned out to be practical for a mixed team. The data engineer set up the connections and tests. A marketing analyst on the social team, comfortable with logic but not with writing large amounts of code, could open the pipeline, see exactly how "qualified view" was calculated, and propose changes through review.
The normalization rules were the heart of it. A qualified view counts only plays above a set watch time on every platform that reports watch time. Engagement rate is calculated against reach everywhere, never followers. Platforms that can't supply a comparable figure show a blank, not an estimate.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
The pipeline also checks itself. If an export arrives with unexpected columns, or a platform's daily numbers drop to zero, the run stops and alerts the team. The three-week shares-and-comments mix-up couldn't happen again silently.
Designing the Dashboard People Actually Read
With clean tables in place, the team used dashboard software to build the view leadership sees each week. The design decisions were mostly about restraint.
The first screen holds five numbers: qualified views, engagement rate, clicks to the website, orders attributed to social, and cost per order for paid posts. Each carries a short definition directly underneath, visible without clicking. Platform comparisons appear only for normalized metrics. Raw platform figures live on a separate page labeled clearly as not comparable.
Annotations mark when big posts went out, when paid campaigns started, and when a platform changed its algorithm or metric definitions. Without them, spikes and dips invite wild theories.
Most dashboard software can refresh constantly, but the team settled on daily updates with a weekly review. Social numbers bounce around hour to hour, and checking them constantly produced more anxiety than insight.
What the Pipeline Can't Tell You
Data showed which videos drove orders. It couldn't explain why. The team still reads comments every day, because a video can perform well on numbers while attracting complaints about an ingredient claim, and that matters more than any chart.
Small samples also mislead. A single viral post can make a whole month's averages meaningless. The weekly review looks at medians alongside averages for exactly that reason.
Four Times the Views, Revisited
On normalized numbers, the platform that once looked four times stronger ranked third of five. The brand moved budget back toward the platform that actually sent buyers to its website, and social-driven orders rose about 30 percent the following quarter.


