The SeekArgus NAF Benchmark · Method

How the SeekArgus NAF Benchmark works

The benchmark puts two scores side by side for each company: how well its story is built, and how the market sees it right now. The gap between the two tells a company what to fix first.

The benchmark runs on a dataset SeekArgus built and maintains itself, so its scores and comparisons appear nowhere else.

What the benchmark measures

The NAF score rates the story a company tells about itself, from 0 to 10. It is the average of four parts:

  • Friction: does the story name a real problem the buyer feels today?
  • Lived Stake: is there a named person with something at stake behind it?
  • Distinctive Qualification: can the company prove things a rival can't?
  • Real Aspiration: does it describe a future only this company could build?

The perception score rates how customers, employees and the market see the company today, from 0 to 100. It is built only from public data that anyone can check.

A strong story with a weak reputation needs a different fix from a strong reputation with a thin story. Putting the two scores together is what tells them apart.

Where the data comes from

Only sources written by someone other than the company count toward the perception score. Everything the company says about itself is kept, but it is used to check the story and the recommendations, never to grade the company.

SourceWhat it showsHow it's used
G2 star ratingWhat customers think of the product, often from smaller teamsCustomer score
Gartner Peer Insights star ratingWhat customers think, often from enterprise buyersCustomer score, averaged with G2
Glassdoor ratings"Recommend to a friend" and "positive business outlook": how employees feel about the job and the company's directionEmployee score
Crunchbase Heat ScoreHow much attention the company is getting in the marketMomentum score
Analyst placementsIndependent recognition from Gartner, Forrester, IDC, Constellation and ChartisProof for Distinctive Qualification. Not in the perception score, because vendors pay for reprints
Press releasesWhat the company says and does right now (Business Wire, PR Newswire, GlobeNewswire)Checks that the story is current, and the prior-work check on every recommendation. Not in the score
News and communitiesWhat outsiders say in their own words (news alerts, practitioner forums, Reddit)Dated evidence with links. In pilot, not yet in the score

The NAF scores themselves come from the SeekArgus leaderboard of 108 companies, each scored against a public source log. That scored set is SeekArgus's own, built analysis by analysis.

How the scores are calculated

Each part of the perception score is put on a 0 to 100 scale, and the three parts are weighted equally.

  • Customer score: a star rating of 3.5 counts as 0 and 5.0 counts as 100. Where both G2 and Gartner Peer Insights have enough reviews, the two are averaged.
  • Employee score: the average of Glassdoor's "recommend to a friend" and "positive business outlook" percentages.
  • Momentum score: the Crunchbase Heat Score, which already runs from 0 to 100.

Minimum data. A review source needs at least 30 reviews to count. A company without enough data on customers, employees and momentum gets no perception score, and is marked Too new to measure with the reason.

The two lines. A company's position depends on two cutoffs:

  • NAF 7.5, the median of all 108 leaderboard scores. Half the companies score above it and half below.
  • Perception 60, a round cutoff we set. It sits close to the median of the 14 companies scored so far, which is 59.7.

Those two lines split the chart into four positions, described below. Every calculation runs on formulas, so any score can be traced back to its inputs.

What the benchmark can't tell you

The benchmark shows where a company stands. It does not prove why.

  • It is a snapshot. Reviews and employee ratings collect over years, so they trail a news event by weeks or months.
  • It shows a link, never a cause. In the first study the rank link between NAF and perception was negative (−0.49) across seven companies. That could easily be chance, and it does not mean a better story hurts reputation.
  • Star ratings bunch up. Most products sit between 4.2 and 4.7 stars, so a gap of 0.1 means little.
  • Some scores sit on a line. A company within one point of 60 is marked borderline, because a handful of new reviews could move it.
  • Employee scores move for many reasons. Leadership changes and layoffs can shift them with little connection to the story.

Reading the positions, and what each one calls for

The position tells you which kind of move comes first. The specific move always comes from the company's own data and its weakest NAF part.

Where a company sits decides which move comes first Running on reputation Write down why customers stay Aligned Keep the proof current Double gap Fix the proof, then the story Under-told Carry the story to where opinions form 60 and up Under 60 Under 7.5 7.5 and up NAF score: how well the story is built Perception: how the company is seen Too new to measure: build early proof buyers can check

The dashed lines are the two cutoffs: NAF 7.5 and perception 60.

PositionWhat it meansThe riskThe kind of moveIn the first study
AlignedThe story and the reputation agreeCoasting. The proof behind the story goes staleKeep the proof current: new numbers, new named customers, a dated update to the founding storyCast AI (8.0, 72.2), MongoDB (7.6, 66.4), Honeycomb (8.0, 60.6, borderline)
Under-toldThe story is strong, but it isn't reaching the people who form opinionsA good story that only lives on the homepageCarry the story to where reviews and ratings are formed: onboarding, support replies, the careers page, employee all-handsNeo4j (8.3, 58.0), Elastic (7.9, 58.7)
Running on reputationPeople think well of the company, but the story doesn't explain whyWhen growth slows, nothing holds the reputation upWrite down why customers stay and where the company is going. Start with the weakest NAF partDatabricks (7.0, 71.3), Snowflake (7.0, 67.1). Both score lowest on Real Aspiration (5.0 and 5.5)
Double gapNeither the story nor the reputation is workingRivals define the company for itFix the proof first, then the story. A story built on weak proof makes the gap worseNone in this study
Too new to measureNot enough public reviews yetThe story is close to the only signal the market hasBuild early, checkable proof: named customers, published results, a founder's account of why the company existsEight companies, including Nebius, Qdrant and Baseten

Reading the parts

Inside each position, the three parts of the perception score point to where the problem lives.

What the data showsWhat it suggestsWhere to actExample
Customers rate the product well, employees rate the company's direction poorlyPeople believe in the product, but the story about the future isn't landing inside the companyReal Aspiration: the careers page, the CEO's internal updates, how leaders explain the product's directionHoneycomb: customer score 73, but 26% of employees have a positive business outlook (45 reviews)
Customers rate the product well, momentum is lowThe product earns its reviews, but the market isn't hearing about itDistinctive Qualification: turn customer results into public proof the press and analysts can useElastic: customer score 67, momentum 42
Momentum is high, customer and employee scores trailAttention is running ahead of experienceFriction: make sure the story names the problem customers actually feel, not only the one investors want to hearNone in this study
What customers praise in reviews differs from the homepage's main claimThe company is selling one thing and being valued for anotherFriction and Distinctive Qualification: build the story on what customers already praiseChecked by comparing the top G2 themes with the homepage line
Too few reviews of any kindThe market has little to go on besides the company's own wordsLived Stake: a named founder or customer gives buyers a person to trustThe eight companies too new to measure

Every example in a published analysis names its figure, its source and the date it was read.

From finding to prescriptive move

A finding only becomes advice after it passes five steps. A move that fails any of them is dropped.

  1. Name the gap with a figure. State the number, where it came from and the date it was read. A gap with no checkable figure behind it is not used.
  2. Tie it to one NAF part. Pick the part the figure points to, usually the company's weakest. That part sets the projected score the move aims to raise.
  3. Write the move in full. Say what to change, where it lives (the page, the post, the asset), the first step, and the metric that will show whether it worked.
  4. Check it hasn't been done. Search the public record for the idea and its synonyms, then look through the company's newsroom, blog, research library, events and the CEO's and CMO's public posts for the last three years.
    • Not found: the move can be published.
    • Partial: the company does something close. Reshape the move so it does what the existing program doesn't, and say how in one sentence.
    • Already done: drop it and write a new one, which goes through the same check.
  5. Publish at least one move for free. Every public analysis gives away at least one complete move, rechecked within seven days of publishing, with the line: "No public evidence the company has done this, as of <date>."

A worked example, using a made-up company. Company X is Under-told: NAF 8.1, perception 57. Its customer score is 71, but only 30% of employees have a positive outlook. That points to Real Aspiration, its weakest part at 6.0.

The move: publish the CEO's three-year product direction on the careers page and open each all-hands with one customer result that shows it working. The first step is a 300-word direction note signed by the CEO. The metric is the Glassdoor outlook figure over the next two quarters. If the prior-work check finds no public version of this, it can be the free move.

The check covers public sources only. It can't see a company's internal plans, so a move is never described as something the company "hasn't thought of."

How often it updates

  • Review and employee data: refreshed each quarter, and again whenever a company is analyzed. The old figures are kept so each company's trend builds over time.
  • NAF scores: set when an analysis is published, and rescored when the company's story changes in a way that matters.
  • News and community evidence: collected weekly once the pilot starts, and kept separate from the score until the samples are large enough to trust. The target is the 40 most recent independent articles per company.

See the leaderboard · See the first industry study

Tim Donovan · SeekArgus