It finds the wasted spend. You see every change.
Worked every night, not every other Tuesday — and nothing moves without a written reason.
✓Sign in with Google. Read-only, no card.
Account audit — last 90 days
Meridian Retail — Search
64/100 · C
The free audit: six areas scored from your last 90 days. Where we start — not what you’re buying.
What a monthly check-in misses
The waste never arrives as a line item. It arrives as CPA drift, one quiet week at a time.
The Tuesday leak
Search terms nobody has read in weeks keep spending every day. Waste doesn’t wait for your reporting cadence.
The retainer math
A base fee plus 10% of spend, for someone who opens the account twice a month. You’re paying for attendance, not for changes.
No change log
Something moved last week. You can’t see what changed, who changed it, or whether it worked.
[ 01 / 06 ]Smart Bidding
You don’t fight Smart Bidding. You feed it.
Google optimizes inside the budget you set, toward the conversion signal you gave it. It does that well, in real time, at a scale no human matches. We don’t try to beat it in the auction.
What it won’t tell you is that the signal is double-counting. That search terms nobody on your team would have bought are still spending every day. That your CPA target is wrong for what these customers are actually worth. Those are structural questions, and they get answered outside the auction.
So that’s the work: harvest the negatives, tighten the match types, set budgets against real pacing, derive the target from margin instead of habit. Then hand Smart Bidding cleaner inputs and let it do the part it’s good at.
Key takeaway
The auction is Google’s job. The inputs are ours.
Applied 04:12
Auto-appliedAdd 3 negative keywords — Brand · Exact
Three search terms spent €412 across 90 days with no conversion. Blocking spend is bounded and reversible.
Queued 04:12
Needs your approvalRaise daily budget €40 → €58 — Search · Non-brand
Raise exceeds the 30% auto-apply ceiling. Campaign cleared the ROAS floor at 3.1× over 30 days.
Two real proposal shapes: one the policy applied, one it held back — and why.
[ 02 / 06 ]The ledger
Every change proposed. Every change reversible.
Nothing moves without a written reason and the policy you set. Bounded fixes apply overnight; anything larger waits in the queue with the numbers that triggered it.
- Propose
- The engine reads the account and writes the change out in full — what moves, from what to what, and the numbers behind it.
- Policy
- Your policy decides. Bounded and low-risk applies itself; everything else queues for you, with the rule it tripped named on the card.
- Apply
- Applied overnight, then re-checked against the policy at apply time — if the account drifted, it demotes itself back to the queue.
- Revert
- One click on the history page puts it back, because the prior state was snapshotted before anything moved.
[ 03 / 06 ]How it works
Connect the account. Approve the first fix.
One Google sign-in, and the audit builds itself from your account’s own data — no forms, no sales call first.
Connect
Sign in with Google and pick the account. Read-only, and you can revoke it the moment you’ve read the report.
Get the audit, free
Six areas scored from 90 days of your own data, with a 30/60/90-day roadmap. Read-only: nothing moves.
Pick who works the queue
Managed, and our operators work it for you. Or Copilot, and you approve the changes yourself.
It runs daily
Bounded fixes apply overnight. Anything larger waits for a human. One click reverts any of it.
Built for the person who answers for the spend
Owner-operated service businesses — home services, legal, dental and clinics — spending their own money on Google Ads.
- Bids, budgets, negatives, ads and structure — changes, not alerts
- A written reason attached to every change
- Revert anything from the history page in one click
- Structure and expansion work, not just upkeep
- Month to month. No contract, no minimum spend
- The same tools over MCP, if you’d rather drive from Claude Code
[ 04 / 06 ]Two modes
Same engine. You choose who works the queue.
Both modes read the same data, propose against the same policy, and log to the same history. You’re choosing how much of it you want to do yourself.
0%
Largest budget step that can apply without you
0
Auto-applied changes per account, per rolling 24h
0×
Hard ceiling on any budget change, re-checked at apply time
0
Changes outside the policy you set
[ 05 / 06 ]Pricing
A base fee, plus a share of what we’re managing.
No plan to outgrow and no cliff when you scale — on Managed the rate steps down as you grow. We read the spend from the account you connect, so your bill matches your Google invoice.
Free audit
Where everyone starts.
$0
No card. However many accounts you connect.
- Six areas scored, with 90 days of data behind them
- A 30/60/90-day roadmap, written for your account
- Every finding shown with the numbers behind it
- A share link and a PDF you can send as-is
- Read-only — nothing changes in your account
Copilot
You work the queue.
$349 /mo
$99 + 1% of $25,000
1.4% of your spend, all in
- Bounded fixes apply nightly, with the reason attached
- Everything larger queues for your approval
- Search-term mining and negative harvesting
- One-click revert on every applied change
- Full MCP access from Claude, Claude Code, or Codex
- Never bills over $499/mo
Managed
We work the queue.
$1,999 /mo
$999 + 4% of the first $50,000, then 3% to $150,000, then 2% above $150,000
8.0% of your spend, all in
- Everything in Copilot
- Our operators work your approval queue
- Structure and expansion work, not just upkeep
- Override any decision we make, from the same queue
- The rate steps down as you scale — 4%, then 3%, then 2%
- No minimum spend — new accounts welcome
Spend is read from the account you connect, so your bill tracks your real Google invoice — nothing self-reported. There is no minimum spend on Managed: the base covers the work whether you're spending nothing yet or scaling hard, and its rate steps down as you grow. Above $250,000/mo we quote it.
[ 06 / 06 ]Why it costs less
The sweeping is software. The judgment is what you pay for.
An agency charges a base fee plus around 10% of spend, and most of that pays for hours: someone pulling search-term reports, checking budgets against pacing, writing up the change log. That work is real — and it’s also the part a machine does better, every day instead of every other Tuesday.
So we split it. The engine does the sweeping: it reads the account continuously, proposes each fix with a reason attached, and applies the bounded ones itself. People do the judgment — what to test, what to scale, when to leave it alone. You pay a base plus 4% instead of a base plus 10%.
Because the base is fixed and the rate is flat, the gap widens as you grow. At $10,000/mo of spend Managed costs about what a retainer costs — the argument there is the nightly cadence, not the price. At $50,000 it’s roughly half. At $100,000 you’re paying about 5% of spend for work that happens every night.
Key takeaway
You’re paying for changes, not for attendance.
The questions that decide it
Stop paying for attendance. Start paying for changes.
Connect the account. The audit is ready in minutes — and it tells you exactly what this would cost you.
Free. Read-only — nothing in the account changes. No credit card, no contract.