Skip to content
LontaAI
Back

Inside

One Tuesday, written the way the log is written

Below is one working day on one account, written the way the log is written — including the checks that found nothing, which is most of them, and the entry where a person overrules the agent.

Entries in this one working day
20
Checks in a full account review
102
Changes that reach you without a named approver
0

What a written list looks like on an ordinary Tuesday

Your account gets a full review every three days against 102 written checks, plus the search-term, pacing, anomaly and call work that runs daily. Below is one working day on one account, written the way the log is written — including the hours where nothing happened.

Checks in a full account review
102Checks in a full account review
How often that review runs
3 daysHow often that review runs
Types of change the system can make
45Types of change the system can make
Agents doing the volume work
50+Agents doing the volume work
Search terms, pacing, anomalies, calls
DailySearch terms, pacing, anomalies, calls
Changes that reach you without a named approver
0Changes that reach you without a named approver

Change log · sample day, not client data

  1. 06:02Agent

    Overnight delivery check. Four campaigns, eleven ad groups, 214 keywords. Spend pacing within 6% of plan, no disapproved ads, no ad group sitting at zero impressions. Nothing to flag.

  2. 06:14Agent

    Offline conversion import ran at 05:40. Nine sold-job records moved from your CRM into Google Ads, none rejected. The import has not skipped a day in 26 days.

  3. 07:31Agent

    Search-term review. 1,842 terms since Friday, 63 of them with three or more clicks. Nineteen read as irrelevant. Each one queued with a note on the demand it would block if we have read it wrong.

  4. 08:06Person

    Two of the nineteen rejected. “ac repair cost” is a price query, and price queries are a buyer segment, not noise. “ac repair company reviews” is somebody checking you out before they call. Three clicks is not evidence either way. Both stay live, flagged for a second look at forty clicks.

  5. 08:20Change

    Seventeen negatives added at campaign level rather than ad group, so the same term cannot come back in through a different group.

    Approved · account lead · 08:19

  6. 08:47Agent

    Month-to-date pacing on all four campaigns. Emergency AC repair at 103% of plan, Furnace repair 96%, Meta 88%, Performance Max 71%. Recommends moving $600 a month out of Performance Max and into Emergency AC repair.

  7. 09:12Person

    Held until Thursday. Budget moves go in late in the week on this account: made on a Monday, the platform spends its re-learning across the whole week and the next read tells us nothing about the change. Recommendation stays on the board, dated.

  8. 09:40Agent

    RSA asset review across six ad groups. Four headlines have sat at Low for thirty days. Six replacements drafted, built from the two same-day angles already rated Best.

  9. 10:05Change

    Four went live as written. Two we rewrote, because both promised same-day service and you do not work Sundays. An ad that gets the click and then disappoints the caller costs you more than the old one did.

    Approved · account lead · 10:05

  10. 10:33Agent

    Landing-page speed regression. /ac-repair, mobile LCP was 2.1s on Friday and 3.8s this morning. Cause is a 1.9 MB PNG uploaded to the hero on Monday afternoon. That page sits behind $4,120 a month of Search spend.

  11. 11:26Change

    Image re-encoded, hero swapped, LCP back to 2.2s on a throttled mobile run at 11:24. The original file is still in your repository, untouched.

    Approved · account lead · 10:58

  12. 11:40Agent

    Fourteen inbound calls transcribed and scored since 17:00 yesterday. Nine qualified, three wrong number, two rang past sixty seconds with nobody picking up. Both of the missed ones landed between 12:40 and 13:10.

  13. 11:52Person

    Sent to you, not fixed. Whoever covers the phone over lunch is not covering it, and there is nothing inside an ad account that repairs that. Two recordings attached. You can keep buying those clicks or you can move somebody's break. Both are decisions, and this one is yours.

  14. 12:15Agent

    Creative fatigue flag, Meta. Spring tune-up offer at 3.4 seven-day frequency against a 2.8 threshold, click-through down 31% over fourteen days, CPM flat within 4%. The audience is not used up. The creative is.

  15. 12:52Change

    Two of five creatives paused, three new statics live on the same offer, same audience, same budget. One variable moved, so next week's number means something.

    Approved · account lead · 12:50

  16. 13:40Agent

    CRM stage sync. Forty-six records touched, six were marked sold overnight, two have sat in Quoted for 21 days against a six-day median. Both came from Emergency AC repair. Flagged to you. No change made in the account.

  17. 14:22Agent

    Anomaly. Cost per click on “emergency ac repair” up 44% between 13:00 and 14:00 against the fourteen-day hourly baseline. Auction insights show one advertiser new since Monday, holding 22% impression share.

  18. 14:40Person

    No action. The auction got more expensive; nothing broke. Bidding back up to win it would cost more per customer than a customer is worth here, and one hour is one hour. Watching for seven days. If it holds, you get a number and a recommendation, not a bid change we already made.

  19. 16:05Agent

    Ran the 22 tracking checks. Is the tag firing, is anything being counted twice, are sold jobs coming back from the CRM, is a 12-second call being counted as a lead. All 22 passed. Fourth week in a row with nothing to fix, which is how it should be.

  20. 17:20Person

    Written into the log at end of day. Four changes live and approved. One budget move held for Thursday. One auction under watch. One problem that is yours and not ours, with the recordings. All of it was already in your account before you read this line.

Sample day, sample account — the checks, the cadence and the approval rule are exactly what runs. Approvers show as roles here; in your own log they are a person's full name.

The division of labour

Agents are good at volume and repetition. The four-thousandth search term gets read as carefully as the first, at 06:00, on a Sunday. They are bad at judgement: an agent will minus a term that was a buyer, or read a two-day dip as a trend, and it is equally confident when it is wrong. People keep budget, strategy, what gets killed, and the approval on every change.

What you can see

The log is not something we send you at the end of the month. It is open in your account the whole time, including the days we changed nothing.

We’ll write the plan for your service area

Six sections within three business days. You do not need an active ad account; if you have one, read-only access is all we ask for.

We do not ask for a card and we do not need a call. Read-only access only, and the findings are yours whether or not you hire us.

Would rather write to a person? [email protected]