How to Vet Influencers: The 100-Point Rubric

What we're actually solving

The operator problem is asymmetric evidence. The creator knows their audience; the buyer sees a profile page and a rate. Most vetting processes fail not because operators cannot spot fake followers but because the check happens ad hoc, differently each time, and the reasoning never gets written down, so the same mistake is re-made each quarter by a different person on the team.

A weighted rubric solves the accountability problem first and the fraud problem second. When the sub-scores are recorded, a bad booking can be traced to the signal that was misread, the weight that was wrong, or the threshold that was too generous, and the rubric gets better instead of the blame getting redistributed. That is also why every number in this rubric is a house rule-of-thumb you are meant to retune against your own outcomes, not an industry constant.

Step by step

1. 01

   Step 1. Score follower quality (25 points)

   Read the audience geography report against the markets the brand sells in, and read the follower growth curve against posting cadence. The house calibration: a majority in-market audience with growth that maps to content scores full marks; an audience that is mostly outside the market, or growth that spikes without matching posts, loses half the points or more. The evidence that moves this score is the audience report itself, which is why the rubric asks for it before anything else. A creator who will not provide a report that is standard for their tier has answered a question before the scoring starts.

   Deliverable

   A 25-point sub-score with the geography percentages and the growth-shape note attached.

2. 02

   Step 2. Score engagement authenticity (25 points)

   Compare the engagement rate to the format average first, because a raw number means nothing without context: per the Collabstr 2026 report (collabstr.com), averages run 6% on YouTube, 5% on Instagram Reels, 4.5% on Shorts, and 2% on TikTok. Then read the last five comment fields by hand. The house calibration: engagement in a sane band around the format norm with specific, arguing, product-aware comments scores high; engagement far above the norm with thin comments, or comment strings of template phrases and bare emoji, loses points fast. Saves and shares count more than likes, because purchased engagement concentrates in likes.

   Deliverable

   A 25-point sub-score with the format comparison and five hand-read comment fields noted.

3. 03

   Step 3. Score audience overlap (20 points)

   Check how much of this account audience already follows creators the brand has booked, from follower-list comparison or the audience report. The house calibration: overlap under roughly a third of the audience scores full marks; overlap past half loses most of the points, because the placement is paying to re-reach people the roster already touches. The evidence that moves the score is the overlap comparison itself, and the reason it gets its own weight is that overlap is invisible in a media kit and expensive to discover after the fact.

   Deliverable

   A 20-point sub-score with the measured overlap share against the current roster.

4. 04

   Step 4. Score content quality (20 points)

   Review the last ten posts for craft and fit: does the opening three seconds earn the watch, is the posting cadence current, does the voice resemble the brand or clash with it. The house calibration: current cadence with varied hooks and a voice adjacent to the brand scores high; dormant accounts, one-format repetition, or a voice that would need heavy art direction loses points. The evidence that moves the score is the posts themselves, and the honest read takes about ten minutes, which is why this signal is skipped more often than any other.

   Deliverable

   A 20-point sub-score with three example posts cited as evidence.

5. 05

   Step 5. Score brand safety (10 points)

   Scan the last ninety days of posts and captions for undisclosed ads, controversies, and anything the brand would not stand beside. The house calibration: a clean recent history with past sponsorships properly disclosed scores full; one undisclosed ad is a deduction and a question, not an auto-decline; anything in the controversy class caps the score at zero and ends the evaluation. Ten points sounds small, but this is the signal that produces the headlines, and it is the cheapest one to check.

   Deliverable

   A 10-point sub-score with the ninety-day scan noted, including anything found.

6. 06

   Step 6. Total the score and act on the decision bands

   Sum the five sub-scores and apply the house bands: 70 and above advances to briefing; 55 to 69 advances conditional on the audience report arriving first; 40 to 54 passes at smaller scope, product-first compensation and a smaller deliverable; below 40 declines, with the reason recorded from the sub-scores. Record the sub-scores, not just the total, because the sub-scores are what make the next vetting round faster and the bands retunable against outcomes.

   Deliverable

   A total score, a band decision, and five recorded sub-scores with evidence notes.

What good looks like: a worked example

A disclosed hypothetical: a home-decor creator with 48,000 followers quotes for a kitchen-goods campaign. Follower quality scores 18 of 25: the audience report shows 71% in the brand market, and growth is steady except two unexplained spikes that keep it from full marks. Engagement authenticity scores 20 of 25: Reels run a 3.8% engagement rate against the 5% Reels average, slightly under, but the comment fields read genuine, with specific questions about products visible in frame. Audience overlap scores 14 of 20: 22% of the audience follows creators already booked, low enough to add new reach. Content quality scores 17 of 20: cadence is current, hooks vary, one caption needed a rewrite. Brand safety scores 9 of 10: one undisclosed ad in the ninety-day window, disclosed after the question was asked. Total: 78 of 100, advance to briefing. This example is a fictional worked case for calibration, not a real creator.

Then the operational layer. In managed work the score is followed by five gates, each a yes that must exist before the deal closes: the location is verified against the declared market, the comments are reviewed by hand, the real quote is confirmed in writing, written usage rights are signed, and payment is agreed after delivery. The rubric and the gates do different jobs, and both are needed: the rubric decides who is worth the process, the gates decide whether a specific deal is safe to close. One house operating record from February 2026 shows what skipping the rubric costs: a creator brief drafted without it re-scored at 29 of 100, and the deal as first priced would have cost $571 per thousand impressions.

Common mistakes

The most common mistake is scoring from the media kit instead of from evidence. A media kit is a sales document; the rubric reads audience reports, public post history, and comment fields. The moment the follower count in the kit becomes the anchor, the rest of the scoring drifts toward justifying it.

The second mistake is running the rubric after the shortlist is emotionally committed. Once the team is excited about a creator, every sub-score gets the benefit of the doubt. The rubric runs before names circulate, and the sub-scores travel with the shortlist so the reasoning is visible to everyone who inherits the decision.

The third mistake is treating the bands as permanent. The thresholds here are house calibration from 2026 campaign work, and they should move when outcomes say so: if 60-score creators keep outperforming the band they were assigned, the band is wrong, not the creators. The record of sub-scores is what makes that retuning possible.

Metrics to track

Rubric coverage: the share of briefed creators that went through the full scorecard. Anything under all of them means the check is being skipped under deadline pressure, which is exactly when it matters most.

Average score of booked creators, tracked over time: if the average climbs, sourcing is learning; if it flatlines while volume grows, the rubric has become a form.

Audience-report collection rate before any figure is quoted: the single strongest predictor of a clean deal, and the easiest discipline to let slide.

Post-campaign outcome by band: whether 70-plus bookings actually outperform 55-to-69 bookings on the metric that mattered. This is the feedback that retunes the bands, and without it the rubric never improves.

Decline reasons recorded: every decline should name the sub-score that drove it. A decline log with no reasons is a process nobody can audit.

Where a planning-first tool fits

Inside Superdirector, the rubric lives as a scorecard on the campaign planning canvas: the five sub-scores, the evidence notes, and the band decision travel with each candidate, so the shortlist conversation starts from the recorded reasoning instead of the media kit. The scoring is done by the operator; the canvas keeps the rubric and the evidence together. The same discipline runs in a spreadsheet, and the spreadsheet version of this rubric is a perfectly good place to start.

Frequently asked questions