Profit Forensics

Give me one week with your ad account. I will find the money it is losing, or prove it is clean. Signed, either way.

A fixed-fee forensic examination of your Google Ads by Igor Ivitskiy, PhD: nineteen years in paid advertising, $770M+ overseen. The verdict is the product.

By application. A few engagements a month.
Instrument  /  The SplitLive
YOUR AD BUDGET buys customers the money working buys nothing the leak, to the dollar scan
What I do
Find the split, to the dollar.
Fee
$3,000 · one week
"Igor could see things that none of the experts I consulted before could."Igor Kheifets · world's highest-paid email marketer
The Meeting

You have seen this meeting.

The report says it was the best quarter yet. The profit says otherwise. Nobody in the room can point to which dollars are wasted.

The agency swears by its ROAS. The CFO calls it fiction. The argument is six months old, and you are expected to be the referee.

You run the ads yourself, evenings and weekends. You suspect part of the budget buys nothing. You have no way to check.

Your team wants a bigger budget. Before you sign it, you want one person with no stake in the answer to look. Everyone you ask has something to sell you.

The Mechanism

Three ways an account flatters its owner.

It grades its own homework.

The platform that spends your money also reports how well it was spent. So it takes credit generously: for your brand name, for returning customers, for people who were coming anyway. I have seen a 99% "optimization score" on an account wasting a fifth of its budget, and a 49% score on the cleanest account in its niche.

reported saleswere cominganyway

One quiet change, months of bleeding.

Bids, budgets and rules change on dates nobody remembers. The damage starts quietly, and the monthly average smooths it into invisibility. The change history names the exact day. I find it.

monthly view hides itthe day it broke

The auction does not know your margins.

It will happily sell you clicks that cost more than the profit they can ever return. And it charges weakness a premium: on one contested term, the weakest player paid three times the leaders' price. The report calls them conversions. Your P&L calls them losses.

click priceprofit a click can returnsold at a loss
What I examine in practice
Search termsthe full term economics, run algorithmically; by hand it is months, so it is never finished
Change historya timeline of causes: which edit fired on which date, and what broke after it
Audiences · geo · hoursevery segment against your margins, not against click averages
Biddingtested against profit per conversion, not the platform's own ROAS
Measurementwhat reported conversions are worth in your books, and where the two disagree
The Examination

One examination, start to finish.

A services business, $40,000 a month on Google Ads, agency-managed. The agency was competent: it hit every target it was given. The report said 4.6x. Here is what the examination said. A redacted engagement, figures rounded.

Found

I put a year of the account's real economics next to what the platform claimed. They disagreed by six figures. That is not an accounting error. That is where an examination begins.

reported 4.6x what the books support 2.4x the gap is six figures a year
Corpus check · 9.5M queries
Not a one-off: mobile takes 72% of budget yet converts worse, not better.
see the finding →
Measured

Behind the gap, four leaks, each with a number and a date from the account's own history: spend flowing to a service line that converts easily but earns thinly ($128k a year), the firm's own clients counted as new acquisitions ($74k), a quiet match-type expansion that kept inquiries flat while quality fell ($61k), and bidding tuned to a number nobody profits from ($49k). Identified: $312,000 a year.

ANNUAL SPEND · ACCT ████ · REDACTED annual spendSPEND −$128kTHIN-MARGIN LINE −$74kOWN CLIENTS AS NEW −$61kMATCH-TYPE CREEP −$49kOFF-HOURS BIDDING $312kIDENTIFIED LEAK $273k of it actionable
Corpus check · 9.5M queries
Across the corpus, the top 1% of search terms carries 62% of conversions.
see the finding →
Verified

Every figure had to survive three independent sources: platform data, analytics, the books. What survived: $273,000 a year, actionable. Every finding in your report names the exact screen in your account where you can see it yourself.

PLATFORMANALYTICSBOOKS Thin-margin service lineOwn clients as newMatch-type creepOff-hours bidding = $273,000 a year, actionable
Corpus check · 9.5M queries
Why verify everything: only 34% of campaigns land within 10% of their own CPA target.
see the finding →
The bar every finding must clear

Three things must hold before anything is priced into the verdict: it costs a measurable amount per year, not "inefficiency" in the abstract; it survives two independent data sources, not one dashboard; and it carries a defined test to confirm or kill it. Whatever fails the three is filed in the appendix, not priced. That is why small stuff never makes the verdict, and why the number that does is one you can defend.

One account is a story. The whole corpus is evidence. The report gives you both.
The Deliverable

What lands on your desk.

One page you will read first: the verdict, the number, the signature. Behind it, the full case: every finding rated by severity, priced per year, tied to a numbered exhibit, and pointing to the screen in your account where you verify it. Then the recovery plan, ordered by dollars returned per hour of work. The verdict answers in three lists: where you lose, where you are clean, and where the evidence supports a scale test.

Executive report · confidentialACCT ████ · redacted
$312,000 IDENTIFIED ANNUAL LEAK Lead mechanism: dated in the account's own change history FINDINGS BY SEVERITY · CROSS-CHECKED ×3 P0Thin-margin service line$128k P0Own clients counted as new$74k P1Match-type expansion$61k P1Off-hours automated bidding$49k NORMALIZED ROAS BRIDGE reported4.6x normalized2.4x CLEAN BILLtracking, geo, budget pacing SignedIgor Ivitskiy, PhDRECOVERABLE · $273k / yr
Redacted engagement, figures rounded.
Verdict
One number, taken to the board.
Findings
Rated, priced, evidenced.
Recovery
Ranked by $/hour, each fix with a test.
Clean bill
If you earned it.
Corpus benchmark
Your pattern vs the category.
Independence, in writing
Signed on page one of every report

Sworn before any finding is delivered

01

I am paid a fixed fee, in full, before any findings.

02

The fee does not depend on what I find.

03

The report is the product. Nothing else is attached to it.

04

I change nothing in your account, and the change history will prove it.

Igor IvitskiyIgor Ivitskiy, PhD · The Google Ads Scientist
SIGNED · FIXED FEE
FORENSIC ARCHIVE
Igor Ivitskiy, PhD
Igor Ivitskiy, PhD · The Google Ads Scientist

The verdict is the value, whatever it says. Owners have paid me to hear "this channel cannot pay for itself, stop." That sentence freed a budget, a team, and a year of attention. Certainty is what you buy. Leaks are just its most common form.

The Engagement
$3,000, fixed.

One account. One week of forensic work. One signed verdict.

The report will either show you where the money is, or prove your account clean. If it does neither, I return every dollar, and you keep the report. A clean bill is not "nothing found": it is a signed verdict that ends the internal argument, and owners have paid exactly for that.

The fee includes a 60-minute findings session after you have read the report. If you want me closer than that, training your team or supervising the fix, that is a separate engagement. Most owners will not need it: the report is built to be executed without me.

Request an examination$3,000 · fixed
Fit is decided by conservative recoverable economics, not spend alone. I check that before payment, and when the report is unlikely to pay for itself, I say so and decline.
Day 0

You request. I ask a few questions, and we both decide it is a fit.

Day 1

A short working call: your goals, your margins, what the account is supposed to do, access.

Day 7

The report is on your desk. The findings session follows when you have read it: the calculations, your team's questions, the validation criteria for every fix.

"That was amazing. I learned more than at another $3k consult."Wayne Yap · $1M aesthetics spa
"In 1 hour, we got ideas to improve our advertising for the year ahead."Mike O'Neil · President, Geek Real Estate Marketing
Straight Answers

The uncomfortable questions, answered.

"You are paid to find problems, so you will find problems."+
The fee is fixed and paid before I look. Drama earns me nothing. A clean bill is a result I sign with pride, and some clients hire me hoping for exactly that.
Why not let my agency audit itself?+
Your agency may be excellent. An independent read exists so the word "may" disappears. Nobody grades their own homework, honestly or not.
Our team is experienced. Why would you find what they cannot?+
Your team manages the account forward. I reconstruct it backward, without delivery pressure, platform incentives, or attachment to past decisions, and I compare its patterns against other industries and account histories. The examination evaluates the account, not the people running it: sound decisions get documented next to the material issues, and that page is your team's to keep.
Is this an AI scan with a nice cover?+
Tools pull the data. The conclusions are mine, checked by hand, and I sign them personally. A scan gives you 200 flags. I give you the few that are worth money, with proof.
Why pay $3,000 when audits are free?+
A free audit is a sales funnel: it has to find work for whoever gives it. You are not paying for hours. You are paying for the absence of a second agenda.
We already keep our negative lists clean.+
So did a clinic running a negative list tens of thousands of lines long, with a third of its spend still going to queries that never converted. Across the corpus, list size and waste are uncorrelated. Hygiene is not diagnosis.
Some of our "waste" is deliberate. How do you tell a strategy from a leak?+
A stated trade-off is recorded as strategy, and its economics are still tested against the goal you set for it. Spend with no documented goal is flagged back to you for clarification before it is ever priced as a leak. Only an evidenced, material gap that serves no goal you recognize is counted as one.
Our primary conversion is a lead or a sample request. Revenue closes later, offline.+
Normal, and the examination is built for it. On the Day 1 call we agree a defensible range for lead value from your own close rates, order values and margins. Verified inputs are cited as verified; uncertain inputs are tested as ranges with their confidence stated. Anything unprovable is marked "not assessable" instead of silently guessed, and if the trail cannot support a verdict at all, I decline before payment.
What if you find nothing?+
Then you get the clean bill and my signature on it, and the internal argument is over. If I can neither find money nor prove the account clean, you get a refund.
My spend is small. Is this worth it?+
Fit is decided by the economics, not the spend number alone: your margins, how measurable the account is, and what a realistic recovery is worth against the fee. I check that at the request stage, and when the report cannot pay for itself, I say so and decline.
What do you need from me?+
Access the way agencies get it, through a manager account. Your margins or a fair estimate, one short call. I change nothing, and the change history will prove it.
Profit Forensics

One week. One verdict. Signed.

Find the money it is losing, or prove it is clean. Either way, you get a report you can act on.