# The Claimed-Customer Grade: Logo Slide to Evidence-Restated ARR

*Grade every claimed customer into one of six evidence-backed grades, restate ARR line by line, and name the accounts that need off-list calls before IC.*

- Canonical URL: https://www.refolk.ai/guides/claimed-customer-grade-arr-diligence
- Pillar: Investing and deal sourcing
- Format: Standard
- Published: 2026-11-14
- Last reviewed: 2026-11-14
- Reading time: 22 min
- Keywords: how to verify startup customer logos, verify ARR claims in due diligence, contracted ARR vs live ARR diligence, check if pitch deck customers are real, pilot counted as ARR red flags, customer list verification before investing

## Key takeaways

- A claimed customer counts as Live Paying only when a bank receipt dated after any opt-out window ties to an invoice and a ledger entry, because HeadSpin's invoices were allegedly altered and an invoice alone proves nothing.
- Under a one-year $120K contract with a three-month termination-for-convenience window, CARR is $120K but ARR is $0 until the window lapses, so that account is Signed Not Live, not Live Paying.
- In published cases, evidenced revenue has run from about 21% of the headline figure (11x) to 90% (Stem), so ARR has to be restated account by account, not with one blended haircut.
- In Refolk's index, 2 of 25 sampled US profiles listing the Gong.io skill (8%) belong to Gong's own staff, so exclude the vendor's domain before treating a product mention as customer evidence.
- The same product shows 265 US profiles and 15 UK profiles in Refolk's index, a 17.7x gap, so a missing public trace for a UK account should be graded Unsupported, not Lapsed.
- Every Unsupported, Lapsed, and Signed Not Live account, plus any account a customer publicly contradicts, goes to off-list verification before the investment memo is signed.

Before IC, give every logo on a startup's customer slide and every row in its data-room customer list a grade that says what the relationship actually is. Then the deck's ARR can be restated on evidence instead of on the founder's word. This standard is for early-stage investors, platform and talent partners at funds, and angels. Its six grades are defined tightly enough that two analysts working from the same deck and data room will assign the same grade. It then turns those grades into a restated ARR table and a list of accounts to verify off-list before the memo is signed.

## What is the claimed-customer grade?

The claimed-customer grade is one label per account, set by documents and cash. It says whether the startup is paid on a live recurring contract, is running a paid or free pilot, holds a signed contract that is not live yet, is still listing a former customer, or is showing a name it cannot support. Every logo and every list row gets exactly one of six grades: Live Paying, Paid Pilot, Free Pilot, Signed Not Live, Lapsed, or Unsupported.

The standard rests on four terms:

- **ARR (annual recurring revenue)** is the annualized value of recurring contracts that are live today. In the strictest treatment it includes only contracted minimum commitments, with no adjustment for overage that has been invoiced or projected.
- **CARR (contracted or committed ARR)** adds signed value that is not live yet. Bessemer describes CARR as adding "committed but not yet live contract values to total ARR and netting out forecasted churn or downsell". According to Founderpath, the SaaS Metrics Standards Board treats committed ARR and contracted ARR as the same thing.
- **Termination for convenience** is a clause that lets the customer exit without cause. The **opt-out window** is the period in which the customer can use it.
- **Off-list verification** is a conversation with someone at the customer whom the founder did not introduce.

GAAP defines neither ARR nor CARR, and neither is audited. No outside standard sits behind the founder's number, so the grade has to come from paper and cash.

### The six grades

| Grade | Required evidence | ARR line |
|---|---|---|
| Live Paying | Recurring contract, go-live passed, opt-out ended, bank receipt dated after opt-out and tied to the ledger | Live recurring |
| Paid Pilot | Pilot or proof-of-concept paper with a fee, plus a receipt for that fee | Pilot, outside ARR |
| Free Pilot | Pilot or trial paper with no fee, or the founder states no fee | Count only |
| Signed Not Live | Executed recurring contract, but go-live ahead, opt-out open, or no qualifying receipt | Contracted, outside ARR |
| Lapsed | Term ended without renewal, or last invoice older than the term | Removed |
| Unsupported | No executed order form, MSA, or pilot agreement in the data room | Removed until verified |

The grades don't overlap because they are assigned in a fixed order, shown in the procedure below. Two tie-breaks settle most disagreements between analysts:

- **A recurring contract with an open opt-out is Signed Not Live, even if cash has arrived.** Revenue Playbook's worked example is a one-year, $120K contract with a three-month unconditional opt-out. Its CARR is $120K and its ARR is $0 for the first three months. Once the opt-out lapses, both are $120K.
- **A cancellable pilot under four months never enters ARR, whatever the paper calls it.** This follows practitioner Aman Samra, who also keeps services out of ARR on a separate line.

> **Rule:** Documents grade, public evidence routes
>
> No account is graded Live Paying on outside-in evidence alone, and no account is graded Lapsed because of public silence alone. Public evidence decides which accounts get calls first. Only contracts, receipts, and the ledger decide the grade.

One gap this standard does not close: I found no published treatment of multi-year prepayment. Until one exists, I grade a prepaid account on its current-year amount and record the prepayment on its own note line. The memo should say this is a house convention, not an industry standard.

## Why do deck ARR and evidenced revenue drift apart?

Deck ARR drifts from evidence because the metric is unaudited and soaks up contracted, trial, pilot, and even former-customer value. In published cases, evidenced revenue has run from about 21% to 90% of the headline. That spread is the reason to grade account by account instead of applying one blended haircut.

| Case | Headline | Evidenced | Share of headline |
|---|---|---|---|
| 11x | ~$14M ARR | ~$3M cleared trial | ~21% |
| Unnamed AI startup | $50M | $42M | 84% |
| One VC's account | CARR = 1.7x ARR | ARR | 59% |
| Stem | CARR | ARR | 90% |

I calculated the share column. The 11x figure comes from a secondary source citing TechCrunch. The 1.7x ratio is one VC telling TechCrunch that CARR ran 70% above ARR at some companies, which is one person's account and not a measured average. Read these rows as the range of what has happened, not as base rates.

Two cases show how the gap gets made:

- **HeadSpin.** The SEC charged the former CEO with an $80M fraud. He allegedly "significantly inflated the value of numerous customer deals and fraudulently treated potential deal amounts that he had discussed with customers as if they were guaranteed future payments", and hid it with fake and altered invoices. The DOJ complaint alleges he told staff to "include revenue from potential customers that inquired but did not engage HeadSpin, from past customers who no longer did business with HeadSpin". ARR was allegedly overstated by about $51M to $55M, and he allegedly "misrepresented that HeadSpin had only lost two customers". A board internal investigation uncovered the scheme. The valuation was revised from $1.1B to $300M, a 73% cut.
- **11x.** TechCrunch reported that one disputed logo, a large sales-data company, described its relationship as a one-month trial, and that Airtable's brief trial "was never used in production". ARR was reportedly counted on full-year contracts that customers could exit after about three months. 11x said it reports contracted ARR and gave its retention as 79%. Former employees described churn of 70 to 80 percent in the first three months. A journalist's inquiries and dated checks of the website brought it to light. One company confirmed Pleo and Rho as genuine customers, so the list was neither all true nor all false.

The common lesson is that both mechanisms leave their trail at the bank, not in the contract. Altered invoices and exitable annual contracts both look fine on paper. In a typical data room, the only document that separates Signed Not Live from Live Paying is a bank receipt dated after the opt-out window. 11x also shows a second pattern: logo walls promote trials.

> A logo can be true on the day it was captured and lapsed by the time IC meets.

## What does each layer of evidence prove?

Each layer of evidence proves less than it appears to. A logo proves a claim. A public trace proves contact. A contract proves commitment. An invoice proves a request for money. Only a bank receipt tied to the ledger proves payment. Grade from the deepest layer you have, and treat everything above it as context.

#### What each evidence layer proves

1. **Claim** - Deck logo slide, website logo wall, data-room list. Proves only that the startup asserts a relationship.
2. **Public trace** - Case studies, staff profiles, awards, listings, archive captures. Proves contact, not payment.
3. **Contract** - Order form or MSA with opt-out, pilot, go-live, and ramp terms. Proves commitment.
4. **Invoice** - Proves the startup asked for money. Can be fabricated or altered.
5. **Receipt and ledger** - Bank receipt tied to an accounting entry. Proves payment.

*Grade from the deepest layer you hold, because every layer above the bank receipt can be true while the customer is not paying.*

### Outside-in sources and how each one lies

- **Customer-side case study**, published by the customer rather than the vendor. It proves the customer was willing to be associated with the product. It lies when it describes an evaluation that has since ended.
- **Customer staff profiles and job posts that name the product.** These prove that people at the customer use or used it. They lie when the profile belongs to the vendor's own staff, or when the product name is a common word.
- **Public contract awards.** These prove a government buyer awarded a contract. Federal agencies must submit contract data within three days of the award transaction, but DoD and USACE data publish with a 90-day delay. A recent defense award can be real and still invisible.
- **Marketplace and integration listings.** These prove the startup built or listed an integration. They lie when a technical partnership is presented as a customer.
- **Archived versions of the logo wall**, from the Internet Archive Wayback Machine. These prove when a logo first appeared and when it was removed. Capture dates are irregular, so a gap between captures proves nothing.

Public-sector awards are the only source with a documented lag. Nobody has published lag or false-positive rates for case studies, marketplace listings, or archived snapshots. Where you rely on them, record the date you captured them and use them to corroborate, never to set the grade.

### How noisy is the product-mention signal?

In Refolk's index of professional profiles, the signal is real but noisy, and how dense it is depends on the vendor and the country. The table below shows three queries and a hand-checked sample from each.

| Query | Profiles | Vendor's own staff in sample | RevOps or sales-ops titles in sample |
|---|---|---|---|
| Gong.io, US | 265 | 2 of 25 (8%) | 5 of 25 (20%) |
| Gong.io, UK | 15 | 0 of 15 | 3 of 15 (20%) |
| Clari, US | 1,299 | 0 of 25 | 1 of 25 (4%) |

There are three readings, each with a check:

1. **Vendor staff inflate the count.** Two of the 25 sampled US Gong.io profiles belong to people who work at Gong. Exclude the vendor's own domain before you count anything.
2. **Short or common-word product names pull in noise.** The Clari US sample included a Limited Partner and the American Heart Association, and only 1 of 25 profiles had a RevOps or sales-ops title. Count a mention only when the product appears next to a relevant title at the named customer.
3. **Density depends on geography.** The US has 17.7x as many Gong.io profiles as the UK, and Clari has 4.9x as many US profiles as Gong.io. Silence is weak evidence against a UK account and stronger evidence against a US account of a widely used product.

**17.7x - US vs UK profiles listing the same product skill in Refolk's index**

265 US profiles against 15 UK profiles for Gong.io, so the same silence means very different things in the two markets.

After those filters, the useful check is narrow: people at the named customer, in a role that would use the product, whose profile names it. [Refolk](/) takes that as a plain-English request across LinkedIn, GitHub, and the open web and returns the matching people. The same search gives you candidate names for the off-list call later.

I ran this search: `Revenue operations or sales operations managers in London whose profiles mention Gong` - [see the full result list](https://www.refolk.ai/s/17yt52j011).

*Returns London RevOps and sales-ops people whose profiles name Gong. Run this shape of query against each claimed customer before you read silence as absence.*

### Data-room documents and the terms that move a grade

The documented diligence trace follows selected accounts from contract to invoice, bank receipt, and accounting entry. Investors ask for a contract registry that covers every active and churned customer, because the churned rows are where Lapsed accounts hide. These terms change a grade:

- Termination for convenience and the end date of any opt-out window
- Pilot terms, including whether a pilot fee was paid
- Start date and go-live date
- Ramp schedule and the current-period amount

Named customer contracts are often released only after the LOI. Plan the grade in two passes. The first pass works from redacted, coded documents. The second confirms names once the LOI is signed.

## How to grade every claimed customer, step by step

Freeze the claims, run the public pass, request and trace the documents, read the terms, then assign grades in a fixed order and send the risky accounts to calls. FE International places data-room review in weeks 1 to 2 of diligence. No other durations for this work are published.

#### The claimed-customer grading procedure

1. **Freeze the claim set** - Capture the logo slide, the website logo wall with dated archive captures, and the data-room list in one sheet. Each row needs a name, a claimed value, and the place the claim appeared.
2. **Run the outside-in pass** - Check for customer-side case studies, public awards, listings, and customer staff profiles or job posts naming the product. Tag each row corroborated, contradicted, or silent, with source and date.
3. **Request the documents per account** - Ask for the order form or MSA, every invoice, and the matching bank receipts, churned accounts included. Record a refusal where documents are withheld.
4. **Trace each account to cash** - Follow contract to invoice to bank receipt to ledger entry. Record the receipt date and ledger reference, or the exact point where the chain breaks.
5. **Read the terms that move a grade** - Pull the opt-out or termination-for-convenience end date, pilot fee, start and go-live dates, and ramp schedule. Record the current-period amount.
6. **Assign the grade and restate ARR** - Apply the decision order below, then build the restated table with live recurring ARR at the current-year rate and contracted, pilot, and services value on separate lines.
7. **Route the calls** - Send every Unsupported, Lapsed, and Signed Not Live account, plus any contradicted account, to off-list verification.

Step 1 matters more than it looks. 11x's logos changed during the period it was being reported on, and a claim set that isn't frozen with dates can't be audited afterwards. Step 6 uses this decision order. Stop at the first question that settles the grade.

#### Decision order for one account

1. **Paper?** - No executed order form, MSA, or pilot agreement in the data room: Unsupported.
2. **Still current?** - Term ended without renewal, pilot ended unconverted, or last invoice older than the term: Lapsed.
3. **Pilot paper?** - Pilot or trial terms: Paid Pilot if a fee receipt exists, Free Pilot if not.
4. **Live and past opt-out?** - Recurring contract with go-live ahead or opt-out still open: Signed Not Live.
5. **Cash tied?** - Bank receipt dated after the opt-out, tied to the ledger: Live Paying. Otherwise Signed Not Live.

*Asking the same five questions in the same order is what lets two analysts reach the same grade from the same documents.*

Use this request for step 3. It asks for the churned accounts and the free pilots explicitly, because those are the rows founders leave out.

**Per-account document request to the founder**

```
Hi {founder name},

Ahead of IC I am restating the customer list account by account. For each account on the attached sheet, could you share:

1. The executed order form or MSA, including any pilot, opt-out, or termination-for-convenience terms
2. The start date, go-live date, and payment schedule, including any ramp
3. Every invoice issued to date
4. The bank receipt that matches each invoice
5. The ledger entry or receivables line for each receipt

Please also include every pilot or trial, paid or free, and every customer that has churned, with the date of the last invoice.

If named contracts can only be shared after an LOI, redacted copies with the customer name masked and a consistent account code across contract, invoice, and receipt are fine for now.

Thanks,
{your name}
```

*Attach the frozen claim sheet and replace the braces. Keep the free-pilot and churned lines, because those are the rows that go missing.*

Sources disagree on when calls and contracts happen. One diligence source says direct customer calls come only at the term-sheet stage. Refolk's guide to customer reference calls says to run the off-list back-channel calls first. Optimist Legal notes that named contracts come after the LOI, which pushes step 4 later. The grade is designed to survive any of these orders: assign provisional grades from whatever you hold, mark them provisional, and finalize when the named documents arrive.

## How to restate ARR from the grades

Restated ARR is the sum of Live Paying accounts at their current-period contracted minimum. Contracted, pilot, and services value each sit on separate lines and never roll into it. The table below is the output IC sees.

| Line | Includes | Basis |
|---|---|---|
| Live recurring ARR | Live Paying | Current-year minimum commitment, no overage |
| Contracted, not live | Signed Not Live | Contract value, shown separately |
| Pilot | Paid Pilot | Fees received, not annualized |
| Free pilot | Free Pilot | Count of accounts only |
| Services | Implementation and consulting | Separate line, never ARR |
| Removed | Lapsed, Unsupported | Zero until verified |

Two rules govern the numbers:

- **Use the current year, not the final year.** Mostly Metrics flags the error of using the final year of a multi-year contract instead of the current year. On a ramped contract, take the amount the payment schedule shows for the current period.
- **Exclude overage.** Under the strictest treatment, only contracted minimum commitments count, with no adjustment for overage invoiced or projected.

If the partnership wants a weighted view as well, Samra suggests haircuts of 0% weight on pilots and 50% on not-yet-live value. Show that as a second column alongside the graded table, never in place of it, and attribute the weights. Pilot conversion data is too inconsistent to support anything more generous:

| Source | Measure | Figure |
|---|---|---|
| IDC/Lenovo study | POCs reaching production | 4 of 33 (12%) |
| IDC CIO Playbook follow-up | POCs reaching production | 46% |
| MIT NANDA | GenAI pilots with measurable impact | 5% |

The two IDC figures are 34 points apart. The MIT figure measures return on investment, not conversion from pilot to paid, so it does not belong in a conversion estimate. No first-renewal loss benchmark for enterprise software has been published either. I give pilots no recurring credit at all.

Each account becomes one row in the grading sheet below. If a second analyst can't reproduce a grade from its row alone, the row is incomplete.

**Per-account grading sheet columns**

```
account, claim_source, claim_location, claimed_value, outside_in_tag, outside_in_source, outside_in_checked_on, contract_type, term_start, term_end, go_live_date, opt_out_end, current_period_amount, last_invoice_date, bank_receipt_date, ledger_ref, grade, arr_line, route_to_call, grader, notes
```

*One row per claimed account. Paste into a spreadsheet as the header row and keep the grader column so disagreements can be traced.*

## How does the grade go wrong?

Most grading failures produce a false Live Paying: a trial, an open opt-out, a ramp's final year, or a fabricated invoice read as proof of payment. The rest misread public evidence, by counting vendor staff as customer staff, treating silence as churn, or leaving a former customer on the list. Each failure below has a check that catches it.

1. **A trial logo graded as Live Paying.** 11x showed a sales-data company's logo that the company itself called a one-month trial. *Check:* require a bank receipt dated after any opt-out window before you grade Live Paying.
2. **An opt-out contract counted at full value.** During the opt-out, CARR is $120K and ARR is $0 in Revenue Playbook's example. 11x reportedly counted full-year contracts that could be exited after about three months. *Check:* read the termination clause and record its end date.
3. **A ramp or exit-year value used as current ARR.** The last year of a ramp is the largest number in the contract and the one most likely to be quoted. *Check:* use the current-period amount from the payment schedule.
4. **Fabricated paper.** HeadSpin's CEO allegedly used fake and altered invoices. An invoice alone is a false positive. *Check:* tie each invoice to a bank receipt and a ledger entry, not to a PDF.
5. **Vendor staff counted as customer staff.** In Refolk's index, 8% of the sampled US Gong.io profiles work at Gong itself. *Check:* exclude the vendor's own domain before you count product mentions.
6. **Short or common-word product names.** The Clari sample turned up a Limited Partner and the American Heart Association. *Check:* count a mention only when the product appears next to a RevOps or sales-ops title at the named customer.
7. **Silence read as absence.** DoD awards publish 90 days late, and small vendors leave thin profile trails, with 15 UK profiles against 265 US. *Check:* grade a silent account without paper as Unsupported, not Lapsed, and send it to a call.
8. **A lapsed customer left on the list.** HeadSpin allegedly counted revenue from past customers and claimed to have lost only two. *Check:* compare the last invoice date with the contract term for every row, including the churned rows in the contract registry.

**8% - Sampled US Gong.io-skill profiles that belong to Gong's own staff**

2 of 25 in Refolk's index, so an unfiltered product-mention count partly measures the vendor, not its customers.

> **Watch out:** Free pilots are invisible to an invoice trace
>
> A free pilot produces no invoice, so a trace that starts from billed accounts never finds one. TechCrunch reported a former employee saying a company "counted at least one substantial, yearlong free pilot as ARR", and the board knew. Start from the claimed list and test each name against receivables, not the other way round.

The second half of that warning is the most common procedural mistake I see. Analysts open the receivables ledger, confirm that each billed customer paid, and call the list verified. That shows billed customers are real. It says nothing about the logos with no bill behind them, and those are the ones that inflate the deck.

## Which accounts go to off-list verification?

Every Unsupported, Lapsed, and Signed Not Live account goes to off-list verification before the memo is signed, and so does any account where public evidence contradicts the claim. Live Paying accounts proven down to the bank need calls only as part of the normal reference sample.

#### Where to spend the call budget

Horizontal axis runs from Paper stops at contract or less to Receipt tied to ledger. Vertical axis runs from Silent or contradicted publicly to Corroborated publicly.

| Quadrant | What it means |
| --- | --- |
| Contact without cash | Chase the receipt. Hold the account at Signed Not Live or Unsupported until cash appears, then call if it does not. |
| Proven | Grade Live Paying. Include in the reference sample only. |
| Highest risk | Off-list call before the memo. These are the rows that sank HeadSpin's and 11x's numbers. |
| Paid but quiet | Grade from documents. Call if contradicted, or if the account is a top revenue line. |

*Accounts with weak paper and no public corroboration take the call budget first, while accounts proven to the bank need only sampling.*

| Grade | Off-list call | What the call settles |
|---|---|---|
| Live Paying | Sample only | Usage and renewal intent |
| Paid Pilot | If counted in deck ARR | Conversion path and date |
| Free Pilot | If shown as a customer | Whether a paid path exists |
| Signed Not Live | Yes | Go-live date and opt-out intent |
| Lapsed | Yes | Why it ended |
| Unsupported | Yes | Whether the relationship exists |

On numbers: one published diligence checklist requires at least three to five independent customer conversations, and at least one back-channel reference that the company did not provide. No published method says how to divide those calls across grades. My policy is to cover every routed account first and then sample Live Paying. If the routed accounts outnumber the call budget, rank them by claimed value.

Finding the right person is usually the slow part, because founder-supplied references are by definition on-list. Procurement, vendor management, and the team that would use the product are the people who know whether an invoice was paid and whether the tool went live. Refolk returns those people at a named customer from a plain-English request, so finding them doesn't depend on the founder's introductions.

## Keeping the grade current until the memo is signed

A grade is dated evidence and goes stale. Re-run the date-sensitive checks right before the memo is signed: opt-out windows that may have lapsed since grading, the archived logo wall, and any defense awards still inside the 90-day publication delay. Upgrades and downgrades both count. A Signed Not Live account whose opt-out ended and whose first post-window receipt has arrived moves up to Live Paying. A logo that has left the startup's own site since grading drops to Lapsed until a receipt says otherwise.

Keep a short change log on the grading sheet: the account, the old grade, the new grade, the evidence, and the date. If IC asks why restated ARR moved between the draft and the final memo, that log is the answer.

#### Before the memo is signed

- [ ] Every logo on the slide, the website wall, and every data-room row appears once in the frozen claim sheet with a source and a date.
- [ ] Every row carries an outside-in tag of corroborated, contradicted, or silent, with the date it was checked.
- [ ] No account is graded Live Paying without a bank receipt dated after its opt-out window and tied to a ledger entry.
- [ ] Every recurring contract has its opt-out end date, go-live date, and current-period amount recorded.
- [ ] The contract registry's churned rows have been checked against the claimed list for Lapsed accounts.
- [ ] The claimed list has been tested against receivables to surface free pilots with no invoice.
- [ ] Live recurring ARR excludes overage, services, pilot fees, and final-year ramp values.
- [ ] Contracted, pilot, free-pilot, services, and removed value each sit on their own line.
- [ ] Every Unsupported, Lapsed, Signed Not Live, and contradicted account has a completed off-list call or is flagged open in the memo.
- [ ] At least one reference was not provided by the company.
- [ ] Opt-out windows, the archived logo wall, and pending defense awards were re-checked within the final week before signing.

## Frequently asked questions

### What is the difference between contracted ARR and live ARR in diligence?

Contracted ARR (CARR) includes signed contracts that are not live yet. Live ARR counts only recurring contracts that are live and can no longer be cancelled for convenience. In Revenue Playbook's example, a one-year $120K contract with a three-month opt-out has CARR of $120K but ARR of $0 until the opt-out ends. GAAP defines neither number and neither is audited, so ask which one the deck is showing, then grade each account yourself.

### Should a paid pilot count as ARR?

No. Put paid pilot fees on their own pilot line at the amount actually received, and don't annualize them. Practitioner Aman Samra keeps cancellable pilots under four months out of ARR and gives pilots a 0% haircut weight. Published pilot-to-production rates range from 12% to 46% depending on the study, which is too wide to support counting any probability of conversion as recurring revenue.

### How do I check if pitch deck customers are real without relying on the founder's references?

Use the outside-in pass. Look for customer-side case studies, staff profiles at the customer that name the product, public contract awards, and archived versions of the startup's logo wall that show when each logo appeared or disappeared. Then call someone at the customer whom the founder did not introduce. Public evidence shows that the two companies had contact. It does not show payment. Only a bank receipt does that.

### What if the startup won't share named customer contracts before the LOI?

This is common, because named contracts are often released only after the LOI. Ask for redacted contracts that carry a consistent account code, along with invoices and bank receipts using the same codes. Grade on those, and mark any account that can't be matched to a name as Unsupported until it can. Record in the memo which grades are still waiting on the named documents.

### Is a missing public trace a red flag for a claimed customer?

Not by itself. In Refolk's index, the same product shows 265 US profiles and only 15 UK profiles. Small vendors and non-US accounts often leave little public trace, and defense contract awards publish with a 90-day delay. Grade a silent account on its documents. If there are no documents, grade it Unsupported, not Lapsed, and send it to an off-list call.

### How many customer reference calls should I make before IC?

One published checklist calls for at least three to five independent customer conversations, with at least one back-channel reference that the company did not provide. No published method says how to divide those calls across grades. My policy is to give calls to Unsupported, Lapsed, Signed Not Live, and contradicted accounts first, and then sample Live Paying accounts.

---

*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/claimed-customer-grade-arr-diligence*
