# Grading a Deck's Traction Claims Against Public Signals

*You will take a deck's stated MRR, user count, and growth rate and grade each plausible, inflated, or unverifiable from public signals in under an hour.*

- Canonical URL: https://www.refolk.ai/guides/grading-deck-traction-claims
- Pillar: Investing and deal sourcing
- Format: Teardown
- Published: 2026-08-11
- Last reviewed: 2026-08-11
- Reading time: 16 min

An investor has a deck in hand and a first call to decide about. This guide is for early-stage investors, platform and talent partners, and angels who need to grade a deck's headline traction before spending an hour on a call. It carries one deck through the whole check, claim by claim, with the real queries, the intermediate counts, and the two flags that looked damning and turned out to be nothing.

The working example is a fictional but typical seed deck. Call it Northstar. The deck claims $5M ARR, "over 100,000 users," and "3x year-over-year growth," positioned as a B2B SaaS company at Series A stage. Those three numbers are what you grade. Everything below is the arithmetic and the searches that turn each one into a written verdict of plausible, inflated, or unverifiable.

## Why the pre-call check is minutes, not hours

The pre-call check exists because the alternative is a wasted meeting, and the meeting is expensive while the check is cheap. Investors spend an average of two to five minutes scanning a deck on first pass, and 80 to 90% of inbound is rejected at that scan. A partner at a mid-tier fund sees 200 to 400 decks a month, which is 10 to 20 hours of screening before a single meeting. DocSend measured average seed-deck review time at three minutes and 44 seconds.

So the pre-call plausibility check is not full diligence. It is a fast, defensible read on three numbers, designed to catch the claims that would embarrass you in the room or waste the hour. The deep unit-economics reconciliation, checking CAC, LTV, cohort retention, and channel concentration against raw source data, belongs after first interest and before a term sheet. At the pre-call stage, steps six and seven below stay light.

**80-90% - Share of inbound decks rejected at the 2-to-5-minute scan**

The screen is measured in minutes, so a written per-claim grade earns its keep through reproducibility, not speed.

The point of writing the grade down is not that it takes long. It is that a second partner can trust your pass or fail without re-running it. A recorded verdict with signal, source, and reasoning is the difference between a screen and a hunch.

## The load-bearing check: ARR per employee

Headcount is a harder number to fake than a traffic screenshot, which is why ARR per employee is the workhorse triangulation ratio. Divide the claimed ARR by a countable full-time-equivalent headcount and compare the result to published medians; a claim that implies an impossibly efficient revenue-per-head lands above the top quartile and needs an explanation.

Here is the band you grade against. Column three is derived: it is $5,000,000 divided by each median, giving the headcount a genuine $5M company at that efficiency would carry.

| ARR band | Median ARR/FTE | Implied FTE for a $5M claim | Source |
|---|---|---|---|
| $1M-$3M | $109,644 | ~46 | saas-capital.com |
| $5M-$10M equity-backed | $152,295 | ~33 | saas-capital.com |
| B2B median (all) | $193,000 | ~26 | getaleph.com |
| B2B top quartile | $279,000 | ~18 | getaleph.com |

The reading is direct. A $5M claim with about 26 employees sits at the B2B median and is unremarkable. The same claim with fewer than about 18 employees lands above the top quartile of $279K per employee, and you should not accept it without a usage-based or AI-native pricing story. The median revenue per employee for private SaaS is $141,125, up from $129,724 the year before, so the B2B median of $193K already reflects a more efficient slice.

For Northstar, the deck says the team is "around 30." Public headcount is where you check that. In Refolk's index, the company shows 28 people with current titles at the firm: 6 account executives, 11 engineers, 4 customer success managers, and the rest split across marketing, ops, and leadership. Twenty-eight people against $5M ARR is $178,571 per employee. That sits between the B2B median and top quartile. Plausible, and consistent with the "around 30" the deck states. First claim survives.

> **Rule:** Count full-time equivalents, not headcount
>
> Divide by FTEs, not raw headcount. Contractors, offshore teams, and outsourced support inflate the ratio if you exclude them and deflate it if the deck quietly includes them. State which you used.

The worked anchors make the sensitivity clear: $5M ARR across 40 employees runs at $125K each, while the same $5M across 20 runs at $250K. A factor-of-two swing in headcount moves you from below median to near top quartile, so getting the count right matters more than any other input in this check.

## The motion decides the proxy

Classify the go-to-market motion before you run any user check, because the motion dictates which proxy is valid and getting it wrong manufactures false red flags. A traffic tool is the right check for a product-led company and the wrong one for a sales-led company; the same low number is a flag in one and noise in the other.

Sales-led growth works when buying decisions involve multiple stakeholders, require customization, and ACV exceeds $25K. Demo-led enterprise models show lower top-of-funnel conversions but significantly higher close rates and larger average deal size downstream. So a sales-led company's website was never the funnel, and its traffic tells you almost nothing about its users.

#### Which proxy is valid for the user claim

Horizontal axis runs from Low web traffic to High web traffic. Vertical axis runs from Sales-led motion to Product-led motion.

| Quadrant | What it means |
| --- | --- |
| Product-led, low traffic | Real flag: PLG needs top-of-funnel, so investigate before grading |
| Product-led, high traffic | Expected: proxy the user claim from traffic, halved for bias |
| Sales-led, low traffic | Noise: the site was never the funnel, use logos and headcount instead |
| Sales-led, high traffic | Bonus signal, but grade the claim on AE count and customer logos |

*The motion decides whether low web traffic is a red flag or expected noise.*

Northstar's deck describes a "sales-assisted" motion with named enterprise logos and an ACV it puts "north of $30K." That is sales-led. So the traffic tool is the wrong proxy for its user claim, and the 6 account executives found in headcount become the primary check. Six quota-carriers closing $30K deals is a modest but coherent sales engine. The "100,000 users" claim, in a sales-led business, almost certainly means seats or free-tier accounts, not paying customers. Flag it not as inflated but as needing definition on the call: users is not the same as customers, and the deck conflates them.

## Reading the hiring footprint

The hiring footprint is the second hard-to-fake signal, and it is where a growth-rate claim gets tested. A company growing revenue 3x should be hiring, and the roles it hires reveal whether the retention infrastructure behind a high-growth claim actually exists.

Public profile data lets you count current headcount, split it by function, and read the last twelve months of hiring. For the pre-call check you want three things: the sales-versus-product split, the pace of recent hiring against the claimed growth, and whether customer success exists at all. A company claiming strong net revenue retention with zero customer success managers is claiming something its org chart does not support.

I ran this search: `Account executives who currently work at HubSpot` - [see the full result list](https://www.refolk.ai/s/24ca0grbas).

*Returns the current quota-carrying sales headcount, the count you divide a claimed sales-led ARR against to test whether the sales engine is real.*

Counting quota-carriers by hand across public profiles is slow and error-prone. [Refolk](/) returns the current holders of a title at a named company in one query, which is what turns a claimed sales-led ARR into a testable ratio. In Refolk's index there are 239,571 people currently holding "Account Executive" titles in the United States, across named SaaS employers including Salesforce, HubSpot, Ironclad, and Liquibase, so the sales headcount at almost any target is countable rather than guessed.

For Northstar, recent hiring shows 4 engineers and 2 AEs added in the last year against a base of 28. That is roughly 20% headcount growth. A company growing revenue 3x while growing headcount 20% is either exceptionally efficient or the growth claim is measured off a tiny base. Both are possible. This is the second flag to investigate before writing anything down.

## The procedure, step by step

Run these eight steps in order. The whole pass is designed to fit under an hour, with the two economics steps kept light at the pre-call stage.

#### Grade three claims in under an hour

1. **Extract the claims** - Pull MRR or ARR, user or customer count, and growth rate verbatim, plus stated stage and motion. Write each as a testable statement with its unit and time period.
2. **Classify the motion** - Decide product-led, sales-led, or hybrid, because it dictates which proxy is valid. Label it so you know whether traffic or headcount is the primary check.
3. **Pull the hiring footprint** - Count current headcount and role mix across AEs, engineers, and customer success. Produce an FTE estimate and a sales-versus-product split.
4. **Compute the ARR-per-FTE band** - Divide claimed ARR by the FTE count and compare to the published band for the claimed size. Note whether the claim sits inside, above, or below the median-to-top-quartile band.
5. **Proxy the user or traffic claim** - Run a traffic tool for web-acquired products; for sales-led, substitute customer logos, review counts, and open job requisitions. Attach a stated error range.
6. **Reconcile unit economics** - Check CAC payback against the ACV band and NRR against the growth claim. Flag each cross-check consistent or inconsistent.
7. **Investigate the damning flags** - Before writing, look for the innocent explanation for your two worst flags. Retire each as a false positive or let it survive.
8. **Write the verdict** - Grade each claim plausible, inflated, or unverifiable with signal, source, and reasoning, so a second reviewer reproduces the grade.

#### What survives each stage of the pre-call screen

| Stage | Figure | Note |
| --- | --- | --- |
| Decks inbound per month | 200-400 | Per partner at a mid-tier fund |
| Survive the 2-to-5-min scan | 20-80 | 80-90% rejected here |
| Get a per-claim plausibility grade | A fraction | This guide's job |

*Volume forces the funnel: most inbound never reaches a verified grade.*

## Reconciling the unit economics

Unit-economics claims fail against the business model, not against a universal target, so pair each claim with the stated deal size or growth rate before judging it. Two cross-checks catch most pre-call problems: CAC payback against ACV, and net revenue retention against growth.

| Metric | Median | Top quartile / target | Source |
|---|---|---|---|
| CAC payback | 16 months | under 6 months | getaleph.com |
| CAC payback, >$100K ACV | 24 months | - | gsquaredcfo.com |
| NRR | 101% | 111%+ | gsquaredcfo.com |
| Magic number | 0.90 | 1.0+ (top >2.0) | rockingweb.com.au |

The CAC-payback row is the sharpest. Companies with ACV over $100,000 show a median CAC payback of 24 months, against 9 months for companies with ACV of $5,000 or less. So a claimed sub-six-month payback on an enterprise ACV is a mismatch, not a triumph. The median SaaS payback is 16 months, down from 18 the prior year, and top-quartile companies recover CAC in under six.

The retention check catches leaky growth. Median NRR has compressed to 101%, with top performers at 111% and above. If NRR is below 100%, every dollar of new ARR is partially offset by churn, so a strong growth claim sitting on sub-100% NRR is leaky and you should demand cohort retention rather than top-line growth. Median growth is around 26%, down from 30% a few years earlier, which makes a "3x" claim worth stress-testing: it is far above the median, so ask what base it is measured from.

For Northstar, the deck claims a "9-month payback" on a $30K-plus ACV. That is roughly consistent with the mid-tier ACV band and does not trip the enterprise mismatch. The deck offers no NRR figure at all, so the retention behind the 3x growth claim is unverifiable from the deck. Note it as a live question.

## Proxying the user and growth claims

Traffic tools proxy web-driven user and growth claims, but they carry a knowable, roughly multiplicative bias, so they read trend better than level. Independent testing found these tools usually exaggerate traffic by two to three times, and that bias is fairly stable per site, so dividing a reading by two gives a usable estimate while the raw number does not.

| Condition | Reliability | Source |
|---|---|---|
| >100K monthly visits | Higher confidence | support.similarweb.com |
| <5,000 monthly visits | Unreliable | promodo.com |
| Typical bias | Exaggerates 2-3x | ivanhoe.pro |
| Direct-tool variance | Up to 30% | support.similarweb.com |

Two thresholds govern trust. Above 100K monthly visits, the tools grow confident; below about 5,000, estimations are not statistically accurate and near-zero readings may be a measurement floor rather than reality. Even direct-measurement tools vary by up to 30% on the same site, so triangulate two sources and report a range.

Because Northstar is sales-led, its traffic proxy is weak by design. Instead, proxy the user claim from customer logos, third-party review counts, and open job requisitions. Named logos you can independently confirm, review counts on public marketplaces, and the pace of open sales roles together give a scale read that a traffic tool cannot for this motion.

> **Watch out:** Never trust a single traffic tool or a near-zero reading
>
> A single tool can be off by 30%, and sites under 5,000 visits cannot be measured at all. Confirm the site clears the floor before you treat low traffic as low usage, and always report a range, not a point.

> Headcount is harder to fake than a screenshot, which is why the arithmetic runs through people, not pixels.

## How this goes wrong: the false positives

The two most damning flags in a pre-call read are often innocent, and investigating them before you write is what separates a defensible grade from a reflexive rejection. Here are the traps and what retires each one.

- **ARR per employee looks too high but is real.** Usage-based companies genuinely lead at about $291K per employee, and AI-native firms run higher than the median band. Check the pricing model and product before calling a high ratio inflated.
- **Low web traffic reads as "no users."** For sales-led enterprise this is normal; the site was never the funnel. Check ACV and whether the motion is demo-led before flagging.
- **Traffic tool shows near-zero.** This may be a measurement floor, not reality, because sites under about 5,000 visits cannot be estimated. Verify the site clears the threshold before trusting the number.
- **Growth looks great, retention hidden.** A high growth claim with sub-100% NRR is leaky. Demand cohort retention rather than accepting top-line growth.
- **Fast CAC payback on enterprise ACV.** Enterprise ACV typically pays back near 24 months, so a sub-six-month claim is a mismatch. Cross-check payback against the stated ACV band.
- **Headcount undercount from contractors.** ARR per employee inflates if the count excludes contractors and offshore staff. Use FTEs, and note which you counted.

Northstar produced two flags. The first: a 3x growth claim against only 20% headcount growth. Investigated, this retired as a false positive. The growth is measured off a small base, and a company can triple revenue off a $1.7M base to $5M while adding a handful of people; the ratio is unusual but not impossible, and the ARR-per-FTE band already confirmed the $5M is plausible. The second flag: "100,000 users" on a sales-led business with 6 AEs. This did not retire cleanly. In a sales-led model, users almost certainly means free seats or trial accounts, not customers, so the claim is not false but is misleadingly framed. Graded unverifiable, with a note to force the customer-versus-user definition on the call.

## Writing a verdict a second reviewer can reproduce

No single canonical public rubric exists for this, so the standard is reproducibility: grade each claim plausible, inflated, or unverifiable, and record the signal, source, and reasoning so a second partner reaches the same verdict. The analyst usually pulls together a written analysis with a recommendation, and independent verification firms issue signed per-claim verdicts; your pre-call version is a lighter, self-serve form of the same discipline.

**Per-claim traction verdict**

```
Claim: [verbatim from deck, with unit and period]
Motion: [product-led / sales-led / hybrid]
Signal used: [ARR/FTE band / traffic proxy / CAC-vs-ACV / NRR-vs-growth]
Public count found: [headcount, traffic estimate with range, review count]
Verdict: [plausible / inflated / unverifiable]
Reasoning: [one sentence a second reviewer could check]
Live question for the call: [what only the founder can resolve]
```

*One block per headline claim. Fill signal and source from your own searches; keep reasoning to one testable sentence.*

Filled in for Northstar, the three verdicts read: $5M ARR, plausible ($178K per employee on 28 FTE, inside the B2B band); 3x growth, plausible but base-dependent (20% headcount growth is consistent off a small base, confirm the base on the call); 100,000 users, unverifiable (sales-led motion means this is likely seats, not customers, define on the call). Two of three survive, one carries a definition risk, and none is a reason to skip the meeting. The written record is what lets a colleague accept that pass in seconds.

#### Before you call the grade done

- [ ] Each of the three headline claims is written as a testable statement with its unit and time period.
- [ ] The go-to-market motion is labelled, and the proxy you used matches it.
- [ ] Headcount is counted as FTEs, and ARR per employee is placed inside, above, or below the published band.
- [ ] Any traffic estimate carries a stated range and the site is confirmed above the 5,000-visit floor.
- [ ] CAC payback is checked against the stated ACV band, and any growth claim is checked against NRR.
- [ ] Both of your worst flags were tested for an innocent explanation and either survived or were retired.
- [ ] Every verdict records signal, source, and one sentence of reasoning a second reviewer could reproduce.

## Keeping the check current

The bands move, so re-anchor them roughly once a year rather than trusting a memorized number. Median revenue per employee rose from $129,724 to $141,125 in a single year, median CAC payback fell from 18 to 16 months, and median NRR compressed to 101%. The direction matters: efficiency benchmarks tighten in good funding climates and loosen in bad ones, so a ratio that read as top-quartile last cycle may be merely median this one.

Two things stay stable enough to lean on. The multiplicative bias in traffic tools, roughly two to three times, is a property of the method, not the year, so the halving heuristic holds. And the load-bearing logic, that headcount is harder to fake than a screenshot, does not decay. When you refresh, update the three tables above from the current benchmark reports, keep the derived FTE column recomputed against whatever the new medians are, and leave the procedure untouched.

The team remains the single most important factor to VCs, cited by 47% in a survey of 885 investors, so remember what this check is for: not to replace judgment on the founders, but to make sure the hour you spend with them is not spent unwinding a number you could have graded in advance.

## Frequently asked questions

### How long should verifying a deck's traction take before a first call?

Aim for under an hour of desk research at the pre-call stage. Investors spend only two to five minutes scanning a deck, and a partner may see 200 to 400 decks a month, so the pre-call check is deliberately light. Reserve the deep unit-economics reconciliation, which marketing-diligence practice puts after first interest and before a term sheet, for later. The pre-call goal is a defensible grade on three headline claims, not full diligence.

### How do you triangulate MRR from headcount?

Count current full-time employees, then divide the claimed ARR by that count and compare to published bands. Median revenue per employee for private SaaS is about $141K, the B2B median is $193K, and the top quartile is about $279K. A $5M ARR claim on fewer than roughly 18 people lands above top quartile and needs a usage-based or AI-native explanation. Use full-time equivalents, not raw headcount, because contractors and offshore teams distort the ratio.

### Why can't I just trust a traffic tool's user estimate?

Traffic tools exaggerate by roughly two to three times in independent tests, and that bias is fairly stable per site, so they read trend better than level. They also cannot measure sites under about 5,000 monthly visits and grow confident only above 100K. Even direct-measurement tools disagree by up to 30% on the same site. Triangulate two sources and report a range, and never trust near-zero traffic before confirming the site clears the measurement floor.

### What makes a CAC payback claim implausible?

A payback claim is implausible when it contradicts the stated deal size. Companies with ACV above $100K show a median CAC payback of 24 months, while those under $5K recover in 9. So a claimed sub-6-month payback on an enterprise ACV is a mismatch worth flagging. Payback fails against the business model, not a universal target, which is why you pair the payback claim with the ACV before judging it.

### What is the difference between an inflated and an unverifiable claim?

Inflated means a public signal contradicts the claim: the implied ARR per employee exceeds top quartile with no usage-based story, or a payback claim conflicts with the ACV. Unverifiable means no public signal can confirm or deny it within the time budget, often because the site is below the traffic floor or the metric is internal-only. Grade honestly: an unverifiable claim is not a red flag, it is a note to test live on the call.

---

*From the Refolk guide library. I revise these guides rather than replacing them, so the current version is always at https://www.refolk.ai/guides/grading-deck-traction-claims*