Leadership and readiness
Supplier application vanity metrics: count what survives three filters
By SourceX Editorial · Reviewed by Noah Loul ·
10 sections
Short answer
Supplier application counts become vanity metrics when they rise faster than the supply they stand for. Normalize them in order: count companies rather than submissions, then keep only those that meet the eligibility rules, have someone with authority approving the discussion and match a dated buyer request. In this article's illustrative year, applications triple while buyer-relevant companies grow 1.5 times and staff hours per buyer-relevant company rise 39.2 percent.
Checklist
Supplier application vanity
Key takeaways
- An application measures interest; supply is an eligible company, with an authorized sponsor, whose records match a dated buyer request.
- Normalize to unique companies first: USCIS counted 470,342 eligible fiscal 2025 H-1B registrations but about 442,000 unique beneficiaries.
- In the illustrative year, only 30 of 400 added applications, 7.5 percent, became buyer-relevant supply.
- Cost per application can fall 30.4 percent while staff hours per buyer-relevant company rise 39.2 percent; track the second.
- Ask for every rung of the ladder, every quarter, under one set of definitions, with rule changes restated.
When do supplier application counts become vanity metrics?
Supplier application counts become vanity metrics when they grow faster than the supply they stand for and nobody shows the gap. An application is a request to be considered. Supply is a company that can license records a buyer wants, and between the two sit duplicates, companies that fail the eligibility rules, applicants with no authority to commit their company and records nobody has asked for.
The decision for an investor is whether a chart of rising applications is evidence that usable supply is rising. Answering it takes records a company either keeps or does not: every submission with its date and channel, the rule that merges submissions into one company, the eligibility rule in force when each was screened, evidence of who approved the discussion, the dated buyer request each company was matched against, and the staff hours spent on all of it, including on companies that went nowhere.
With those records, application growth can be normalized: restated as the number of eligible, authorized, buyer-relevant companies, per period, under one set of definitions. Without them, a tripling of applications is a statement about marketing reach. The sections below set out the unit rule, the three filters, an illustrative year in which the two numbers part ways, and the questions that expose the difference.
Count companies, not submissions: the unit rule comes first
Normalize to unique companies before anything else, because a submission count rewards the same company for applying twice. Public programs that allocate scarce places among applicants publish counting rules, and those rules show what a disciplined unit looks like.
U.S. immigration authorities now run the H-1B cap lottery on people, not filings. Under the beneficiary-centric process described on USCIS's registration page, each unique beneficiary with a registration is entered into the selection once, and a prospective petitioner that submits more than one registration for the same person has all of them invalidated. USCIS's figures show why the unit matters: for fiscal 2025 it counted 470,342 eligible registrations, 47,314 of them for beneficiaries with more than one eligible registration, and about 442,000 unique eligible beneficiaries, an average of 1.06 registrations each. For fiscal 2026 the agency reported about 339,000 unique beneficiaries and 1.01 registrations each.
Research funders apply the same discipline to success rates. NIH's definition counts multiple submissions of the same project in one fiscal year only once, and leaves out applications withdrawn before review, returned or administratively withdrawn. A data company should be able to state its equivalent rule: what makes two submissions the same company (legal entity, corporate group, web domain), which date a merged record keeps, and whether earlier quarters were restated when the rule changed.
Three filters between a company and usable supply
After the unit rule, three filters separate companies that could become supply from companies that merely asked. Each needs a written definition, a dated record and a denominator that does not change mid-series.
- Eligible: the company meets the published rules on the day it is screened. SourceX's live pages, for example, say it focuses on companies with 50 or more full-time employees at peak and set country and language conditions; the count should name the rule version applied to each company, so a rule change does not masquerade as growth.
- Authorized: someone who can commit the company has approved the discussion. SourceX's screening page asks for an authorized sponsor, an owner, executive or designee able to approve a discussion about licensing the company's own data; an application from a curious employee is not yet supply, however promising the records sound.
- Buyer-relevant: the company's records match a dated, written request from a buyer, not a category someone might want one day. SourceX lists Acquisition Partner demand among its fit signals and checks relevance to its current data needs during review; an investor should ask which request each counted company was matched against, and on what date.
- Each filter is a share of the one before it, so the series reads as a ladder: applications, unique companies, eligible, authorized, buyer-relevant. Report every rung for every period, not only the top and the bottom.
The application growth normalizer: one illustrative year
In an illustrative year built for this article, applications triple from 200 to 600 a quarter while eligible, authorized, buyer-relevant companies grow from 60 to 90, a 1.5-fold rise. Every figure comes from a script and stated assumptions; none describes SourceX or any real company.
The ladder narrows faster as volume grows. In the first quarter, 30.0 of every 100 applications became buyer-relevant companies; by the fourth, 15.0 did. Staff hours, charged at half an hour of triage per application, an hour and a half of eligibility checking per unique company, three hours of authority checking per eligible company and four hours of buyer-fit review per authorized company, rise from 19.1 to 26.6 per buyer-relevant company, an increase of 39.2 percent.
The figure plots each rung as an index of its first-quarter value. Read it from the top line down: the further a rung falls below the applications line, the more of the reported growth never became supply. The quarter-by-quarter counts:
| Quarter (illustrative) | Applications / unique companies | Eligible / authorized | Buyer-relevant (per 100 applications) | Staff hours per buyer-relevant company |
|---|---|---|---|---|
| Q1 | 200 / 190 | 120 / 100 | 60 (30.0) | 19.1 |
| Q2 | 260 / 240 | 145 / 118 | 70 (26.9) | 20.0 |
| Q3 | 380 / 330 | 180 / 140 | 80 (21.1) | 22.3 |
| Q4 | 600 / 480 | 230 / 170 | 90 (15.0) | 26.6 |
| Q4 against Q1 | 3.00x / 2.53x | 1.92x / 1.70x | 1.50x | up 39.2% |
Where 400 extra applications went
Of the 400 applications added between the first and fourth quarters, 30, or 7.5 percent, became buyer-relevant supply. The rest stopped at a rung where each still cost staff time, and the bridge below shows where.
Duplicates and ineligible companies account for nearly three quarters of the added volume, 27.5 and 45.0 percent. In the model that is the signature of growth bought through broad campaigns or incentives that reward submissions rather than suitable companies, and it is the easiest kind to fix: tighter targeting, a duplicate check at entry and eligibility questions on the first screen. The authority and buyer-relevance rungs are harder, because they depend on conversations and on buyers' requests, which no form can speed up. The bridge:
| Rung where the added applications stopped | Added, Q1 to Q4 | Share of the 400 added | What to ask the company |
|---|---|---|---|
| Duplicate submissions from the same company group | 110 | 27.5% | What merges two submissions, and were earlier quarters restated under that rule? |
| Unique companies that fail eligibility | 180 | 45.0% | Which rule version screened each company, and did any rule change during the year? |
| Eligible companies with no one authorized yet | 40 | 10.0% | Who approved the discussion at each counted company, and on what date? |
| Authorized companies matching no buyer request | 40 | 10.0% | Which dated request was each company matched against? |
| Eligible, authorized, buyer-relevant companies | 30 | 7.5% | How many of these later reached a signed deal or an accepted delivery? |
| All added applications | 400 | 100.0% | Show every rung for every quarter, under one set of definitions |
Strongest alternative: more applications only raise the cost of supply
The strongest case against celebrating application growth is that more applications can raise what supply costs without adding any. The illustrative year shows how a cost metric can hide it: staff hours per application fall 30.4 percent from the first quarter to the fourth, because many new applications die cheaply at triage, while hours per buyer-relevant company rise 39.2 percent.
Cost per application is therefore a second vanity ratio. It improves precisely when the funnel fills with submissions that never needed real work, which is the condition an investor should worry about. Cost per buyer-relevant company, and later cost per accepted dataset, move the other way and tell the truth about operating leverage.
Application growth does mean supply growth in one situation: when every rung grows at roughly the same rate, so the per-100 conversion holds steady as volume rises. A company that can show a flat or rising conversion from applications to buyer-relevant companies, quarter by quarter and channel by channel, has earned the right to talk about applications. One that shows only the top of the ladder has not.
An application is nonbinding interest, not supply
An application costs the applicant nothing and commits it to nothing, so it measures interest, the same way a nonbinding pre-order does. Presenting interest as if it were demand or supply can mislead investors, as one SEC case shows.
In February 2024 the SEC announced settled charges against Lordstown Motors over statements about demand for its electric pickup, finding that the company had touted nonbinding pre-orders from customers who in many cases did not operate fleets or plan to buy the truck. The case concerned a vehicle maker's pre-orders, not data supply, but the mechanism carries over: a count of nonbinding expressions of interest says little about the transactions that will follow.
SourceX's own application page is explicit on this point. It tells applicants that no files, passwords or API tokens are needed and that nothing is licensed or transferred by applying, and its how-it-works page says that qualifying does not assure an offer. Those statements are accurate descriptions of what an application is, and they are also the reason an investor should never read application counts as supply.
What could make a normalized count mislead as well
A normalized count is better than a raw one, but each rung can drift or be gamed. Four risks deserve a question in diligence.
- Rules move. Widening eligibility rules, adding countries or relaxing a size threshold lifts the eligible count without any change in who applied. Each period should be restated under the current rule, or reported under both.
- Buyer relevance expires. A company counted as buyer-relevant against last quarter's requests may match nothing this quarter. Buyer-relevant supply should be dated to the request, and stale matches reported separately.
- Authority lapses. Leaders change, companies are acquired and sponsors leave. An authorization recorded a year ago may not survive the first contract conversation.
- Buyer-relevant is still not accepted. Rights review, preparation and a buyer's own acceptance test come later and cut the count again; normalized intake is a leading indicator, not delivered supply.
Application metric questions for any data company, SourceX included
Ask for the ladder, the rules behind each rung and the hours, and apply the same questions to SourceX.
SourceX's live pages describe an application of three screens and about seven minutes, in which a company names its systems, how many years of history they hold and what it can export; SourceX then reviews fit, finishes the inventory with the company, checks rights and sets a processing route. SourceX publishes no application, conversion or supply counts; those numbers, like its other company and commercial metrics, go to investors on request. SourceX is raising money under Rule 506(c), a route that admits accredited investors alone, and it answers requests for materials through its investors page. Questions worth putting to any data company:
- What is your unit rule for merging submissions into one company, and when did it last change?
- For each of the last four quarters, how many applications, unique companies, eligible, authorized and buyer-relevant companies did you record?
- Which eligibility rule version applies to each period, and are earlier periods restated?
- What evidence shows who approved the discussion at each authorized company?
- Which dated buyer requests define buyer-relevant, and how many counted companies match a request older than two quarters?
- What were staff hours per buyer-relevant company and per accepted dataset in each quarter?
Methods note: the normalizer's inputs, rules and limits
The application growth normalizer, its bridge and the cost comparison were built for this article on 2026-10-28 as an editorial model, computed by a short script (xs0187_application_normalizer.py) kept with its output in the writer's working folder. Every input is illustrative and none is a SourceX figure. The quarterly counts are set in the script (applications 200, 260, 380 and 600; unique companies 190, 240, 330 and 480; eligible 120, 145, 180 and 230; authorized 100, 118, 140 and 170; buyer-relevant 60, 70, 80 and 90), and the script checks that each rung is no larger than the one before it.
Staff hours are half an hour per application, one and a half hours per unique company, three hours per eligible company and four hours per authorized company, summed per quarter. The bridge subtracts each rung's first-quarter gap to the next rung from its fourth-quarter gap, and the five parts sum to the 400 added applications. Index values divide each quarter by the first quarter; buyer-relevant companies per 100 applications divide the two counts; hours per application and per buyer-relevant company divide the quarter's hours by each count; growth multiples and percent changes compare the fourth quarter with the first. Arithmetic uses exact fractions, and values are rounded once, half up, at output, so changes are computed from unrounded values.
Public evidence: USCIS's H-1B electronic registration page and NIH's success rate definition, both checked through search excerpts of their own sites; the SEC's February 2024 press release on Lordstown Motors, checked the same way and paraphrased; and SourceX's how-it-works, application, investors and screening pages fetched on 2026-10-28. Not measured: any real company's application volume, conversion or hours, at SourceX or elsewhere; overlap between channels; how long authorization or buyer relevance lasts; and how many buyer-relevant companies later reach acceptance.
Frequently asked questions
What is a vanity metric in a data supply business?
A number that rises when the business looks busier but not when it gets better at producing usable supply. Raw application counts qualify when duplicates, ineligible companies and unauthorized applicants grow faster than companies a buyer could use. The test is whether the normalized count grows at the same rate.
How should a data company deduplicate supplier applications?
Merge submissions that belong to the same legal entity or corporate group, using registered names, domains and contacts, and keep one dated record per company. State the rule, apply it to every period, and restate earlier quarters when it changes, as public programs such as the H-1B lottery do when they count people rather than filings.
Is cost per application a useful efficiency metric?
Not on its own. It falls when many cheap, unsuitable submissions enter the funnel, which is exactly when real cost rises. Track staff hours per buyer-relevant company and per accepted dataset beside it.
When does application growth mean real supply growth?
When conversion from applications to eligible, authorized, buyer-relevant companies holds steady or improves as volume rises, quarter by quarter and channel by channel. If conversion falls, the extra applications are diluting the funnel rather than adding supply.
Does SourceX publish its application or supplier counts?
No. Its live pages describe how applications work and say that applying licenses nothing, but it publishes no application, conversion or supply figures. It shares company and commercial metrics with investors directly; this article's numbers are illustrative.
Sources
- USCIS describes a beneficiary-centric H-1B selection in which each unique beneficiary is entered once and duplicate registrations by one petitioner are invalidated, and reports 470,342 eligible fiscal 2025 registrations, 47,314 for beneficiaries with multiple registrations, about 442,000 unique beneficiaries (1.06 registrations each) and about 339,000 for fiscal 2026 (1.01 each). Source
- NIH's success rate definition counts multiple submissions of the same project in one fiscal year once and excludes applications withdrawn before review, returned or administratively withdrawn. Source
- In February 2024 the SEC announced settled charges against Lordstown Motors over statements about demand for its pickup that relied on nonbinding pre-orders, many from customers that did not operate fleets or plan to buy the truck. Source
- SourceX's how-it-works page describes its intake (systems, years of history, what can be exported, then fit, inventory, rights check and processing route), its focus on companies with 50 or more full-time employees at peak, its fit signals including Acquisition Partner demand, and that qualifying does not assure an offer. Source
- SourceX's application page describes three screens of about seven minutes and says no files, passwords or API tokens are needed and nothing is licensed or transferred by applying. Source
- SourceX's investors page says it shares company and commercial metrics with investors directly and offers investor materials on request. Source
- SourceX's screening page lists an authorized sponsor able to approve a licensing discussion and relevance to current data needs, which SourceX checks during review, among its screening criteria. Source
Request SourceX investor materials
SourceX shares company and commercial metrics with investors directly. Any offering is made only to verified accredited investors.