The trend chart is the dashboard's most-looked-at element and its most misread one. It plots your visibility over time — each point is the share of runs that mentioned your brand, across the prompts, platforms and date range you're viewing.
Narrowing what the line covers
Unfiltered, the chart is one account-wide average, and averages hide the story. Filtering is what makes it readable:
By tag — compare product lines, markets, or funnel stages against each other instead of reading one blended number. This is the payoff for tagging your prompts.
By platform — strength on one assistant and absence from another cancel out in the blend. Filtering to a single platform is also the only fair way to compare periods around the date you added a platform.
Choosing a date range
Day-to-day movement is mostly variance — assistants answer the same question differently between runs, and that's how the models behave, not a signal. Movement over weeks, across many runs, is signal. When the daily line looks jumpy, widening the range is usually the honest fix rather than a way of hiding something.
One edge worth knowing: the most recent point can still be filling in while checks are running. Reading today's dip at 9am is reading a half-finished day.
Why the line bends when nothing happened
Three edits move the chart without anything changing in the market:
Adding or archiving prompts. The line averages whatever is tracked, so archiving prompts you were losing on raises it and adding hard new ones lowers it. If the line moves sharply on a day you edited your prompt list, you're looking at your own edit.
Adding a platform. Same mechanic — the blend now includes somewhere you're stronger or weaker.
Adding an alias. Different mechanic, bigger effect: aliases recount your history, so the whole line can shift, past included. That's the line repairing itself, not a glitch.
Comparing two periods honestly
Three rules. Compare like-for-like lengths — a week against a week, a month against a month. Hold the prompt set fixed across both periods, or you're partly measuring your own editing. And when a comparison matters — a campaign, a content push — note the dates you changed prompts, platforms or aliases, because those are the bends you'll otherwise misattribute to the work.
If the chart looks flat when you expected movement, Why your data hasn't updated yet covers the usual reasons — most often the number of new runs is simply small against the history it's averaged into.
