White-label · for agencies and web consultants

Add a code-level performance audit to your agency — without building the team.

Your brand. Your price. Our evidence engine. Every PageSpeed warning judged with evidence — the published Noise & Silence method — and every deliverable reaches your client under your identity.

Your wholesale cost: $298.20 per audit
Public list price: $497 40% below it
Your retail price: You decide

No monthly minimums
No seat fees
2 verification re-runs included
Delivered in less than 24 hours
Agencies worldwide

Inspect the deliverable under your brand ↓

The same report, two identities Standard report Your agency

PageSpeedAudit Noise & Silence audit Performance audit

Worked example · the client's page · mobile + desktop

4 fixes worth shipping

17 warnings owed no work

81% Noise Rate

1 Silent Bottleneck

Silent Bottleneck field p75 TTFB 1.6 s — passing in the lab, in none of the 21 warnings

Fix 1 · render-blocking chain · Real Google's "save 2,330 ms" replaced by a range recomputed against the field p75 LCP of 2,784 ms — arithmetic shown.

Prepared by PageSpeedAudit · pagespeedaudit.com

Your agency Performance audit Performance audit

Worked example · the client's page · mobile + desktop

4 fixes worth shipping

17 warnings owed no work

81% Noise Rate

1 Silent Bottleneck

Silent Bottleneck field p75 TTFB 1.6 s — passing in the lab, in none of the 21 warnings

Fix 1 · render-blocking chain · Real Google's "save 2,330 ms" replaced by a range recomputed against the field p75 LCP of 2,784 ms — arithmetic shown.

Prepared by Your agency

An illustrative deliverable built from the framework's published worked example — not a client site. Only the identity changes. Inspect the worked example →

Inspect the deliverable

Put your name on something you have actually seen.

Your client judges your expertise through this report. Switch between the four artifacts, then read the complete worked example they are drawn from — every row below is quoted from it.

One engagement, in the order your client meets it 01 Executive summaryFor the person payingOn delivery 02 Evidence reportFor the developer checkingThe same day, in their reply 03 Implementation planFor the team shippingThe sprint that follows 04 Verification re-runsFor proving the resultOnce the fix ships

01 · Executive summary

Your agency Executive summary Worked example

What is slowing the page — and what is not

4 fixes worth shipping

17 warnings owed no work

1 bottleneck the warnings never named

The decision. Do not treat the 21 PageSpeed warnings as a backlog. 4 are Real. 6 are Wrong, 9 Trivial, 2 Locked — each with a documented permission to ignore it, an 81% Noise Rate. A separate real-user signal, a field p75 TTFB of 1.6 s that passes in the lab, is the binding constraint, and it appears in none of the 21.

The summary states what was measured, what was decided and what remains uncertain. It does not promise revenue; the Speed → Revenue calculator prices the opportunity separately, as an evidence-bounded range.

02 · Evidence report

Your agency Full evidence report 21 warnings

Every warning receives a verdict — with the evidence quoted beside it

PSI flag	Evidence consulted	Verdict	Rationale
Reduce render-blocking resources — “save 2,330 ms”	Field p75 LCP = 2,784 ms	Real (impact recomputed)	The chain is real — but the saving is a lab-graph counterfactual, and subtracting it from the field p75 (implying a ~450 ms LCP) is arithmetic across incompatible baselines; honest range recomputed against field data.
Properly size images	srcset / sizes present and correct; flag fired on 2x density at emulated DPR	Wrong	Correct responsive art direction misread as oversizing.
Ensure text remains visible during webfont load	Icon font with font-display: block	Wrong	Blocking is the correct behaviour for icon fonts; swap would flash raw ligature text.
Image elements do not have explicit width and height	Lab CLS 0.122; field p75 CLS 0.00 with an essentially all-good CLS distribution	Trivial	The shift appears under cold-cache lab conditions; the field distribution shows no materially affected cohort.
Reduce unused JavaScript — 909 KiB	The flag's own metricSavings: zero on the metrics modelled; bytes attributed to reCAPTCHA	Trivial	Zero modelled savings and security-owned, interaction-path code — the verdict rests on the combination, never on the zero alone.
font-display — “est. savings 10 ms”	Savings an order of magnitude below single-run variance	Trivial	Below the instrument's own noise floor.
Serve static assets with an efficient cache policy	Assets served by the platform's CDN; TTLs not merchant-controllable on this plan	Locked	True and material — and owned by the platform; effort redirected to the controllable equivalent.

Seven of the worked example's rows, quoted from the framework. A client report carries the full table with verbatim evidence per row, a documented falsification attempt for each verdict, and a confidence grade — no artifact, no verdict.

03 · Implementation plan

Your agency Implementation plan Fix 1 of 4

Fix 1 · the render-blocking chain

HTMLstylesheethero image

What changes: The stylesheet the HTML discovers before the hero image comes off the critical path — the corrected template block, verbatim for the client's platform, with the file or location it belongs in and the lifecycle constraint that makes it safe to paste.
The prediction: Not Google's "save 2,330 ms". A range recomputed against the real users' p75 LCP of 2,784 ms, with the arithmetic shown — subtracting the lab number would leave a ~450 ms LCP, arithmetic across incompatible baselines.
The proof: 2 re-runs within 30 days measure the page again once the fix ships. The comparison on your dashboard reports, finding by finding, whether the problem is still detected and how the field metrics moved against a stated noise floor — with the predicted range beside each for you to judge.

Fixes 2–4 follow the same shape, in impact-per-effort order; the 17 non-actions are listed with the reason each may be ignored. Want to test a change before anyone ships it? The Fix Simulator (free account) predicts it first.

04 · Verification re-runs

Your agency Verification Predict → prove → confirm

The prediction is re-measured, not repeated

Predict Every fix ships with a predicted range against field data and the arithmetic behind it.
Prove After the client's team ships, the included re-runs measure the page again — 2 within 30 days. The comparison on your dashboard reports, finding by finding, whether the problem is still detected, judges field metrics against a stated noise floor, and marks a lone lab-score move as what it is: not proof.
Confirm Real users decide last. CrUX aggregates a rolling 28-day window, so a real improvement surfaces as a four-week ramp; the free Core Web Vitals watch emails the day-28 verdict.

Two clocks, started by two different events

The second clock starts when the client's team deploys, not when the audit does. Ship on day 25 of a 30-day window and the re-run still shows the structural fix — but real-user data will not have refilled by the time the entitlement closes. That is the instrument's clock, not a failed fix, and the free Core Web Vitals watch carries the observation past the window as a separate service with its own scope.

Do the predictions hold? In the published calibration pilot — 8 controlled cases — the realised saving was never smaller than the estimate: 1.0–1.9× the prediction. Read the pilot and its limits →

The same decision trail as JSON — schema pagespeed-audit/findings@1, excerpt with the worked example's values

{
  "schema": "pagespeed-audit/findings@1",
  "run": {
    "binding_constraint": "field p75 TTFB 1.6 s — passing in the lab, absent from all 21 flags",
    "adjudication_counts": { "confirmed": 4, "false_positive": 6, "deprioritized": 9, "platform_blocked": 2 }
  },
  "adjudications": [
    {
      "audit_id": "render-blocking-resources",
      "disposition": "confirmed",
      "claim": "save 2,330 ms",
      "evidence": "Field p75 LCP = 2,784 ms; the lab saving is a counterfactual on a different baseline — impact recomputed against field data"
    }
  ]
}

Your identity changes

Agency name, logo, accent colour, tagline, website, contact email and footer note — set once in your account, applied to every audit you order.

The audit substance never does

Evidence, verdicts, priorities, prediction arithmetic, implementation detail and the verification method are the same on every report, whoever's name is on it.

Read the complete worked example → See the delivery pipeline → See exactly what every buyer gets →

Why the client's developer accepts it

Warnings and real-user evidence enter through different doors.

An agency's risk with any white-label supplier is the moment the client's own developer reads the report. This is what survives that conversation.

Lane 1 · Google's warning list

21 warnings, equal apparent urgency

4 act on · 17 owed no engineering time — an 81% Noise Rate

Lane 2 · real-user field data

What the list never names

p75 TTFB 1.6 s Needs improvement

Not the worst band — and still the binding constraint for this page. Those are two different findings, and the report states them separately.

+1 Silent Bottleneck — the binding field constraint, absent from all 21 warnings

One prioritised plan

Both streams, one order of work

Fix 1 The render-blocking chain · Real
Fixes 2–4 The other Real warnings, ordered by impact per effort
Then Server response time · the Silent Bottleneck. An investigation, not a template change — it is owned by the host or the application tier, so it is scoped separately and does not carry a fix number.
17 Non-actions, each with its written reason

21 warnings adjudicated · 4 real issues to fix · 1 constraint to investigate. Three counts, because they answer three different questions.

Real: Passes the truth, materiality and jurisdiction tests — evidenced, material, and the client's to fix.
Wrong: Fails the truth test: the warning misreads the live behaviour.
Trivial: Fails the materiality test: true, but below the instrument's own noise or the client's decision threshold.
Locked: Fails the jurisdiction test: true and material, but owned by the platform or a vendor.

Inspect the complete Noise & Silence framework → Why a green lab score can still hide a real-user failure → Test a proposed fix before anyone ships it →

The economics

Your cost is fixed. Model the offer you actually intend to sell.

You buy at 40% below the public list price and sell at whatever your market bears. Arithmetic, not a projection: gross profit before your own sales, support, tax and implementation costs.

What will your client pay? $497 $997 $1,500 $3,000

Or any price

…and sell this many a month 1 2 5 10

Include my own sales and delivery time

Sales hours per audit Delivery hours per audit Your hourly cost

Client pays: $497
Your wholesale cost: −$298.20
Gross profit per audit: $198.80
Gross margin: 40%

$198.80 gross profit a month at 1 audit

$397.60 gross profit a month at 2 audits

$994 gross profit a month at 5 audits

$1,988 gross profit a month at 10 audits

Client pays: $997
Your wholesale cost: −$298.20
Gross profit per audit: $698.80
Gross margin: 70.1%

$698.80 gross profit a month at 1 audit

$1,397.60 gross profit a month at 2 audits

$3,494 gross profit a month at 5 audits

$6,988 gross profit a month at 10 audits

Client pays: $1,500
Your wholesale cost: −$298.20
Gross profit per audit: $1,201.80
Gross margin: 80.1%

$1,201.80 gross profit a month at 1 audit

$2,403.60 gross profit a month at 2 audits

$6,009 gross profit a month at 5 audits

$12,018 gross profit a month at 10 audits

Client pays: $3,000
Your wholesale cost: −$298.20
Gross profit per audit: $2,701.80
Gross margin: 90.1%

$2,701.80 gross profit a month at 1 audit

$5,403.60 gross profit a month at 2 audits

$13,509 gross profit a month at 5 audits

$27,018 gross profit a month at 10 audits

What one audit leaves you

— contribution margin

Client pays—

Wholesale cost—

Contribution—

A month at ——

Break-even client price with the costs you entered—

Contribution after the costs entered above — your wholesale cost and the hours you priced. It is not profit: tax, tooling and everything this page cannot see are still to come.

Enter a client price to see what the audit leaves you.

At this price the audit costs more than it earns — raise the price, or narrow the surrounding work.

Under a 25% contribution margin, one refund or one long client call can erase the spread — price the service around the report, not the report alone.

Compare every scenario

Gross profit a month by client price and volume, at today's $298.20 wholesale cost — before any time of yours, because these rows cannot know your hours. Your own row is the exception: it carries the hours you entered above, so it reports contribution rather than gross. The two bases are not interchangeable, which is why the row says so.

Client price	1 / month	2 / month	5 / month	10 / month	Gross margin
— your price, after your hours	—	—	—	—	—
$497 (list)	$198.80	$397.60	$994	$1,988	40%
$997	$698.80	$1,397.60	$3,494	$6,988	70.1%
$1,500	$1,201.80	$2,403.60	$6,009	$12,018	80.1%
$3,000	$2,701.80	$5,403.60	$13,509	$27,018	90.1%

What the market bears. A comparable agency engagement runs $3,000–$8,000 and takes 2–4 weeks (the comparison on the pricing page); the last preset is the low end of that range. A higher price should buy a wider service — your presentation, a prioritisation workshop, managing the implementation — not the same report with a larger number on it. Your cost moves only if the public price does: the discount is a percentage, not a fixed figure.

Inside your reseller account

The product does not stop at putting your logo on a report.

Your workspace finds the first client in your own public work, starts the proposal from evidence rather than a blank page, and turns unspent audits and expiring re-runs into a reason to call.

Likely client opportunities — from your own public pages

Private to you · we never contact your clients

Businesses your site presents as clients, each anchored to a sentence quoted verbatim from your own case studies or portfolio — no quote, no entry. Under each one: a proposal section and a client email, prefilled with your brand and the client's name.

A client your own pages name

Found on your case-study page — the sentence that names them, quoted

Your client-facing pack — proposal, email, objections Proposal section Client email Objection handling

Proposal section Drop into your proposal template as the diagnostics phase.

Performance audit — evidence before remediation

We begin with a code-level performance audit delivered under our brand. Its method — published as the Noise & Silence framework — adjudicates every PageSpeed warning with evidence: Real for this site, Wrong, Trivial, or Locked by the platform. The deliverable is a verdict on each warning, the deep findings, an executive summary, and the underlying data, delivered in less than 24 hours from commissioning.

The audit includes 3 audit runs within 30 days, which we use to verify each fix against a fresh measurement rather than assuming it worked.

Deliverables:
  - A verdict on every PageSpeed warning, with the evidence for it
  - The prioritised fix list, in your platform's vocabulary
  - An executive summary written for stakeholders
  - Verification re-runs to confirm the fixes landed

Client email Ready to forward. Adjust the greeting; the substance stands alone.

Subject: Proposal — a performance audit before we touch anything

Before we commit budget to speed work, I want evidence about which of the warnings on your site actually matter. We commission a code-level audit for exactly this: every PageSpeed warning is adjudicated — Real, Wrong, Trivial, or platform-Locked — so the work we then quote is limited to fixes that will move your numbers.

The audit is delivered under our brand, delivered in less than 24 hours, and includes verification re-runs so each fix is confirmed against a fresh measurement. I will walk you through the findings when it lands.

If you are happy to proceed, reply and I will commission it this week.

Objection handling The three you will actually hear, with answers that hold.

"We already ran PageSpeed Insights."
So has everyone — that is the list, not the verdict. The audit adjudicates each item on that list with evidence, under the published Noise & Silence method. The deliverable is which warnings to fix, which to ignore, and why: the part the free tool does not do.

"Can our developers not do this themselves?"
They can implement the fixes — that is exactly what the report equips them to do. What it replaces is the two weeks they would spend arguing about which warnings matter, and it settles those arguments with measurements instead of opinions.

"How current will it be?"
delivered in less than 24 hours, from commissioning to the full report. It also includes 3 audit runs within 30 days, so after your team ships fixes, the improvement is verified against a fresh measurement rather than assumed.

Generated from today's facts by the same code that builds an approved reseller's pack; the pack rebuilds itself if the price or the terms change, so nothing you send a client goes stale.

Your offer pack

Matched to the live pricing

Your margin, worked throughNot for publishing
Proposal sectionClient-ready
Client emailClient-ready
Objection handlingClient-ready
When to sell it — and when not toNot for publishing

Your clients

Your book, not a receipt pile

Unspent credits — they never expire; the 30-day clock starts at the first run
Clients audited, and what each audit found
Re-runs left on each audit, and its window
Re-runs expiring within 7 days — the natural reason to call that client back

1 · Find the first client

Opportunities come from the work your own site already shows, not from a purchased list.

2 · Start from evidence-backed material

Proposal, email and objection answers carry your brand and the current offer; you adapt the voice and send it.

3 · Create legitimate repeat contact

Unspent audits and expiring re-runs sit in one view — operational inventory, not receipts.

Free instruments for a discovery call agency reference

Use the same evidence standard before you ask a client to buy. Each tool answers a different question; none replaces the audit. Every one of these opens under our branding with a buy-it-yourself call to action — so they are yours to work from, not material to forward to a client. Opening them in a new tab keeps your place; it does not solve the commercial conflict. What your client is meant to read is the branded proposal and report you send them.

Value the speed An evidence-bounded revenue range, backed by the cost of a millisecond Test a fix Predicted versus measured, before anyone ships (free account) Set a budget Turn an LCP target into a constraint, backed by the cost of a kilobyte Check integrity Is the site serving PageSpeed a faster page than its users? Framed by how many sites cheat PageSpeed — a pilot that found none in 1,162 measured sites, with the wider protocol published Prove a change How many runs a before/after claim actually needs

Explore all five free instruments → Evidence: the cost of a millisecond → Evidence: the cost of a kilobyte → Evidence: how many runs prove a change →

Application to verification

Your client relationship stays yours. The evidence work stays ours.

Who does what, stage by stage — and where we remain invisible to your client.

Your client sees

Your agency

The price you set, on your invoice
Your brand on every page and every report — name, logo, accent, tagline, footer
Your contact details as the only way to reply
The proposal and report at your own address, once you point a domain here

Only you see

Your $298.20 wholesale cost and your margin
Your pipeline, notes and what each engagement earned
The workspace, the sales kit and everything you draft in it
Whether and when the links you sent were opened

We handle, unnamed

Producing the audit and rendering it under your identity
Storing the deliverable and serving it at the address you send
Nothing outbound to your client: we never contact them

The one place to be deliberate: our public research and free tools carry our branding and a direct buy-it-yourself call to action. They are excellent agency reference and they are not client-safe material — the list further down says which is which.

Stage	Your client	Your agency	PageSpeedAudit
1 · Apply	—	Confirm your email; brand once — name, logo, accent, tagline, website, contact, footer.	Reads your site, approves an agency on the spot, unlocks the 40% wholesale rate.
2 · Sell	Agrees scope and price with you. Never registers with us.	Qualifies with the free tools, packages any surrounding service, sets the fee, invoices.	Invisible. The offer pack and the discovery tools are yours to use.
3 · Order	—	Buys at wholesale — checkout or API — and points the credit at one URL.	Runs the full method: lab and field data, every warning adjudicated, the Silent Bottleneck, platform-native fixes.
4 · Deliver	Receives your branded executive summary, evidence report and implementation plan.	Presents the decision, coordinates the work, remains the one they call.	Returns the report and its JSON under your identity, in less than 24 hours.
5 · Verify	Sees, through you, which findings resolved and how real-user metrics moved — with the prediction beside each.	Starts the 2 included re-runs once fixes ship; closes the loop.	Runs the same method again and compares run to run on your dashboard: structure first, field metrics against a noise floor, lab moves flagged as not proof.

After approval, the lifecycle can be scripted.

An API key with the purchase scope orders audits at your wholesale rate — a saved card is charged directly, otherwise the API hands back a Stripe checkout link — then assigns URLs, starts re-runs and collects the branded report and its JSON. Applying is the one browser-only step.

Open the API reference

Reseller or affiliate?

One question decides which program fits.

Neither is better. They differ in who owns the customer — and everything else follows from that.

Do you want to own and support the client relationship? Yes No

When the answer is yes

Reseller this page

Your client buys from you, at whatever you charge, and calls you
The report carries your identity — they never see us
You earn the spread again on every repeat audit
You front $298.20 per audit and answer for the work

Apply to resell

When the answer is no

Affiliate

You publish or send a tracked link; the customer buys from us at $497
You earn $149.10 (30%) on their first audit
Nothing out of pocket, nothing to deliver or support
A confirmed email registration stays yours for 90 days, with no cookie needed

See the affiliate program

A strong reseller fit

Web design and development studios
SEO and digital-marketing agencies
Performance consultancies
Freelance web professionals with active client work — a portfolio site is enough to verify
Teams willing to hand a client an honest "your page is already fast"

Use another path when…

You only need an audit for your own site — buy one directly
You want a monthly seat-based monitoring platform
You want to resell a raw PageSpeed export with a logo on it
You would rather not own the client conversation after delivery — refer instead

Compare every reseller and affiliate detail

Compare	Reseller this page	Affiliate
Your client buys from	You, at whatever you charge	Us, at the $497 list price
The report is branded	Yours — they never see us	Ours
You earn	The spread. Costs you $298.20; you set the price	$149.10 commission (30%)
On their second audit	You earn again, every time	Nothing — commission is the first purchase only
You get paid	When your client pays you	After a 30-day refund hold, once you pass $300
Out of pocket	$298.20 per audit, paid to us	Nothing, ever
The work per sale	You scope it, present it and answer for it	You publish a link
Getting started	Confirm your email, then the website check	Confirm your email — active immediately
Can you do both?	Yes — refer the readers you will never work with, resell to the clients you will. A wholesale order is a wholesale order: it does not also pay an affiliate commission.

Operational details

The questions that remain once the product is clear

Exactly what branding appears on the report?

Your agency name, logo, accent colour, tagline, website, contact email and a footer note — set once in your account and applied to every audit you order afterwards. Soon after approval, the name, colour and tagline your homepage states are read in as a starting point; override any of them.

Does the client ever see PageSpeedAudit?

Not on the report: the branding you set replaces ours on every audit you order. Contractually you remain our only customer and the one responsible to your client — section 10 of the terms says so in plain words.

What exactly does an audit cost me, and are there minimums?

$298.20 at today's $497 public price — 40% below list, a percentage, so your cost moves only if the public price does. No minimums, no seats, no monthly fees: order when you have a client who needs one, with the same 2 verification re-runs within 30 days and delivery in less than 24 hours as every audit.

How is my agency verified, and what if I apply from a personal address?

After you confirm your email we read the website behind your domain for evidence that you serve clients; an agency site is approved on the spot. From a personal address we cannot tell either way, so we email you for your agency's website — reply with one link and it gets the same automatic check. If a site is declined and you think that is a mistake, contact us and a person takes a look.

What happens if a report is wrong?

Every prediction carries its arithmetic and is tested by the included re-runs, and the method is published in full, so a client's developer can check any verdict against the evidence quoted beside it. The commercial promise is report or money back: if we cannot deliver your report, that purchase is refunded — the refund terms and the warranty terms set out exactly what that covers.

Can I audit a client's account area or staging site behind a login?

Yes, as long as the page is reachable over the internet. When you enter the URL, give access with a test account or a session cookie from the client — we sign in once to confirm it works before the run, store it encrypted and delete it when the audit window ends. Ask the client for a dedicated test user rather than their own login. Pages behind a browser password pop-up (HTTP authentication) aren't supported yet. The report is marked as a signed-in audit, and screenshots can be blurred when the page shows customer data.

Can the whole lifecycle run through the API?

Everything after the application: an API key with the purchase scope orders audits at your wholesale rate, assigns URLs, gives signed-in access for pages behind a login, starts re-runs and retrieves the branded report and its JSON findings. Being an approved reseller changes the price you pay and the branding on the report, not what the API lets you do — every account can create a key at /account/api-keys. See the API reference →

Which research do I hand a client who challenges the method?

Three more the thread of a discovery call tends to reach, each with its scope stated rather than dressed as settled: Do WordPress speed plugins work in the field? — a published causal design and replication package, not an effect estimate — for "we already installed a speed plugin". Earlier paint, or less work? for "our INP improved, so the experience improved": part of the web's INP win is repainting sooner rather than doing less, and the study registers a design to size that share. Do soft navigations repay the slow first load? for the SPA argument — a source-derived model, with its comparative benchmark not yet run.

When does lab predict field? for a green score that hides a real-user failure; How many Lighthouse runs prove a change? for "we ran it twice and it improved"; Do Lighthouse's promised savings materialize? for whether predictions hold; The cost of a millisecond for what speed is worth. All published with their data.

Ready when you are

Make the audit part of your agency.

One email confirmation, one website check, one branding setup. After that, every wholesale audit returns under your identity — in less than 24 hours, with its re-runs.

Apply to resell Inspect the sample again

Not ready to apply? Get the three-email reseller playbook.

Email 1The margin math — 40% below list, your retail, and why the spread exists.
Email 2The pitch that survives your client's developer — selling verdicts, not tool exports.
Email 3Setup, approval and your first order — branding once, every report ships as yours.