To measure social media ROI without trusting the platforms' own scorecards, watch one number they can't inflate: how many people search your brand name in Google. Organic social, PR, podcast appearances, and every other awareness play share the same downstream effect. Someone sees you, remembers you, and later types your name into a search box. Google Search Console records every one of those searches, free, and one filter separates them from everything else.
The whole setup takes about ten minutes. What you get is a trend line that answers the question your reach metrics keep dodging: is any of this turning into demand?
Why likes can't answer the question
Platform analytics measure attention inside the platform, and the platform grades its own homework. Impressions, reach, engagement rate: all of it can climb for months while zero new people ever go looking for your business. The feed rewards content that keeps people in the feed. Your business needs content that makes people leave it.
Branded search is the counter-metric. It's downstream (nobody searches your name by accident), it's independent (Google doesn't care how your reel performed), and it's commercial (someone searching "narsil creative" is much closer to a call than someone double-tapping a post). Likes are applause. Branded search is people walking into the shop.
Step 1: build the branded filter
A plain text filter on your brand name undercounts, because people don't search tidily. They run the words together, they hyphenate, they type your domain, they misspell you. A regex catches the family:
(?i)(narsil creative|narsilcreative|narsil-creative|narsilcreative\.com|narsilcreative com|narsul)
Swap in your own brand's variants. We built a free Branded Regex Generator that does this in under a minute: type your brand name and URL, toggle the variants, copy the pattern. It runs in your browser with no signup, and it flags the judgment call that matters (broad single words that could catch strangers stay off unless you turn them on).
Step 2: apply it and compare periods
In Search Console, open Performance, then Search results. Add filter, Query, Custom (regex), Matches regex, paste. Then click the date filter, choose More, and set Compare last 3 months to previous period.
That's the whole mechanism. The branded clicks and impressions columns now show this quarter against last quarter, and the delta is your answer.
Step 3: read it like an adult
Branded search up meaningfully while your posting cadence held: the off-site work is building real awareness. Keep going, and note which weeks spiked against what you published.
Flat or drifting down after months of consistent posting: the content isn't landing, and it's better to know that now than after another quarter of effort. Change the message, the format, or the channel before changing the budget.
And a candid note before you celebrate or panic: branded search has confounders. A press hit, a paid brand campaign, seasonality, even a local news mention will move it. Read the trend against your actual activity calendar, not in a vacuum. The signal is real; it's just not private property of your social strategy.
What a real read looks like
A shaped example, typical of what we see in client accounts. A home-services company starts posting three short videos a week in March. The branded comparison at the end of May:
Branded clicks: 142 vs 61 (previous period)
Branded impressions: 1,890 vs 830
Better than doubled. But the useful part is the week view: two visible spikes, one aligned with a video that got picked up by a local Facebook group, one with a podcast appearance that had nothing to do with the video program. The read isn't "social doubled our brand demand." It's "the video program plus one earned-media hit doubled it, and the group pickup mattered more than the other eleven videos combined." That's a finding you can act on: make more of the thing that traveled.
Expect lag, too. Awareness converts to search over weeks, not hours. Someone sees three of your videos before your name sticks, and searches it the day they need you. Judge a program on a quarter, not a launch week.
Log it or you'll argue about it later
The trend line only becomes evidence when it sits next to what you actually did. One row a month, copy the header into a spreadsheet:
Month | Branded clicks | Branded impressions | Posts shipped | Earned media (PR, podcasts, pickups) | Paid brand running? | Notes
Three months of rows and the "is social working" conversation stops being a feelings debate. Six months and you can see which content formats precede the bumps. This table is also the exact artifact to bring to any agency conversation, including one with us: a vendor who can't engage with your branded-search history isn't measuring anything real.
The bonus view: your SEO without the brand noise
Flip the same filter to "Doesn't match regex" and you get the inverse: every query where someone found you without knowing your name. That's your true non-branded SEO performance, the clean baseline most reports blur by mixing brand demand into organic totals.
The non-branded view is also where the rest of the free toolkit runs: the Search Console AI filter prompts sort those queries by buyer intent, and the striking distance method turns the almost-ranking ones into fixes. Same data source, three different questions.
Who shouldn't bother yet
Two groups. Brand-new sites have no baseline: a domain in its first few months will show single-digit branded queries, and a trend line needs history before it means anything. Come back after a quarter of data. And businesses doing no off-site marketing at all have nothing to measure: this is a thermometer, not a heater. (If your curiosity is about AI assistants rather than Google, measuring your AI traffic is the parallel exercise.)
For everyone in between, the loop is: generate the regex once, check the comparison monthly, log what you shipped that month next to the number. Three months of that and you'll know more about your social program's real effect than any platform dashboard will ever tell you.
We run this exact branded versus non-branded split in every monthly client report, as part of the Assay phase of the Reforge Method, because the first question a measurement layer has to answer is which half of your traffic you earned and which half you already owned.