Home / Blog / From spreadsheet hell to structured analysis
PPC Operations  ·  May 25, 2026  ·  7 min read

From spreadsheet hell to structured analysis: how to automate the reading of your campaigns.

The honest, unglamorous truth about paid media operations in 2026 is that most teams still spend the morning consolidating numbers and filtering tabs. The analyst day is the invisible bottleneck — and it's not the optimization that's broken, it's the reading. Here is what changes when Claude reads the campaigns instead of the operator.

L
Leonardo RochaBuilding actcenter while operating 39 client accounts at RMC PPC

Walk into any PPC team between nine and noon and you will see the same scene. Three tabs open in Google Ads, two in Looker, one in a shared Sheet that someone last updated yesterday. Coffee. A search-term export. A "let me just check pacing real quick" that takes 45 minutes. By the time the strategist has a clean view of the portfolio, half the morning is gone and the part of the job that requires a brain — deciding what to do — gets squeezed into the afternoon.

For years I treated this as the price of doing the work seriously. If you want to know the account, you have to crawl through the numbers. That is what good operators do.

I was wrong. The crawling is the bottleneck. And it is not even the worst part — the worst part is that two analysts crawling the same account on the same day will surface different things, miss different things, and frame the same numbers differently. The work is slow and inconsistent. We just stopped noticing because it has always been that way.

What manual analysis actually costs

The cost everyone names first is time. A solo PPC pro burns 9 to 12 hours a week assembling the view of their accounts before they get to decide anything. An agency strategist with 18 accounts burns closer to 20. The math is uncomfortable once you write it down: a third to a half of a senior salary spent on the work of looking.

But time is the cheap cost. The expensive one is consistency.

When analysis lives in spreadsheets and tribal knowledge, the same account gets read differently by different people. Different cuts. Different windows. Different definitions of "pacing healthy." The senior on the team has their checklist in their head; the junior has theirs in a Notion doc; the contractor has theirs nowhere. Three reasonable people look at the same account and produce three reasonable, different stories. The decisions that flow from those stories diverge in ways that are hard to debug after the fact.

And then there is the person dependency. When your best analyst goes on vacation, the account quality drops for a week. Not because nobody else is competent — because the method lived in their head. You weren't running a process. You were running a person.

The invisible bottleneck

You can hire your way out of a time problem. You cannot hire your way out of a consistency problem. The second one compounds quietly — a slightly different read of pacing here, a slightly different framing of search terms there, until the portfolio's decision quality is governed by who happened to look at which account that morning.

The shift: from filtering tabs to asking questions

The thing that changed for us was not the AI. It was the data. Once the Google Ads scripts started writing structured tabs to a Sheet on a schedule — pacing, conversions, search terms, ad-level metrics, structural drift — the reading job stopped being a manual job. Claude can read those tabs in seconds. The strategist's day inverts.

Instead of opening Google Ads and asking the platform to show them something, the team opens the briefcase and asks a question:

The answer is back in seconds, with the numbers attached and the reasoning shown. No filtering. No vlookup. No "let me just pull a fresh export." The strategist's morning becomes a conversation with the data instead of a wrestling match with the spreadsheet.

That sounds like a productivity story. It is also a consistency story, and the consistency story is the bigger one. We will get to that.

8am 9am 10am 11am 12pm 1pm 2pm 3pm 4pm 5pm 6pm BEFORE consolidate sheets filter campaigns build the report actual decisions warm-up AFTER briefcase decisions · approvals · applies strategic work · client · escalations
Figure 1 — The analyst day, before and after. Same hours on the clock. The block of work that actually requires a brain expands from three hours to nine.

Six tasks that should never be manual again

These are the six pieces of reading work that, once you have structured data and a model that can read it, you will never want to do by hand again. Each one is something we used to do in a tab, and each one is something the briefcase now does before the strategist sits down.

  1. Reading per-campaign performance. The "what changed this week" pass across every campaign in every account. Five minutes per account by hand, twenty seconds in a structured pull. Done before coffee.
  2. Anomaly detection. Spend spikes, conversion drops, CTR collapses, search-term volume shifts. The kind of thing a careful human catches eventually and a tired human misses. Structured comparisons surface these the morning of, not the week after.
  3. Period-over-period comparison. WoW, MoM, YoY across every metric that matters. The work that used to require building pivot tables now lives in a tab that already exists. You ask, you get the answer.
  4. Pacing analysis. Budget pacing across the portfolio, sorted by risk. Who is going to overspend by Friday. Who is underspending and leaving demand on the table. One read, whole portfolio, no spreadsheet.
  5. Search-term evaluation. The single most expensive manual job in PPC. Reading thousands of search terms a week to find the ones to negate and the ones to harvest. Structured data plus a model that can read intent collapses this from hours to minutes.
  6. Cross-signal prioritization. The hardest one — combining pacing, CPA drift, search-term quality, and Quality Score into "these three accounts need attention today." This is the work that only seniors used to do, because only seniors could hold all four signals in their head at once. With structured inputs, the model holds them. The senior decides what to do about it.

None of these six are exotic. They are the things every PPC operator does, every week, by hand. They are also the things that — once moved into structured analysis — collapse from "the morning" to "the first fifteen minutes."

The consistency dividend nobody talks about

This is the part most people miss when they hear "AI analyzes your campaigns." They hear the speed and miss the structure.

When analysis is manual, every analyst on the team is running their own subtly different process. The senior's pacing check is not the junior's pacing check. The Wednesday review is not the Friday review. The way you read search terms when you are fresh is not the way you read them at 4pm. The work is human, which means it is variable, which means the decisions that flow from it are variable.

When analysis is structured, the process is fixed. Same windows. Same definitions. Same thresholds. Same signals weighed the same way every morning across every account. Two strategists looking at the same briefcase get the same diagnosis. The senior on vacation does not break the process.

The dividend is counterintuitive: decision quality does not just go up because the analysis is faster. It goes up because the analysis is the same every time. When you do make a bad call, it is a bad call against a known baseline. You can find the bad call, debug it, and adjust the baseline. With manual analysis, you cannot — every bad call sits inside a slightly different process and there is nothing stable to fix.

Manual analysis

  • Different analyst, different read
  • Different day, different read
  • Errors are random
  • Process lives in heads
  • Senior on vacation = quality dip

Structured analysis

  • Same read, same window, every time
  • Errors are systematic — and fixable
  • Process lives in the briefcase
  • Vacation-resilient
  • Decision quality rises with seniority of the process, not the person

I find myself caring about this differential more than the time differential, even though the time differential is what closes the demo. The hours saved are real. The hours saved compound when the work that fills them is more consistent than the work it replaced.

What you keep doing by hand (and why that's fine)

If I left the post here, it would read like a pitch for "automate everything." That is not the bet. There is a long list of things the strategist still does by hand, and they should.

The decision of whether to scale a winner or stabilize a loser is a human call. The conversation with a client whose pacing is off because their phone team stopped answering is a human call. Writing the email that explains why we are pulling spend on a campaign the client loves emotionally is a human call. Creative direction is a human call. Choosing which negative to push to the shared list versus the account-level list is, sometimes, a human call. Reading whether a search term is "expensive low-intent" or "patient research mode" is a human call.

What changes with structured analysis is not that humans stop deciding. It is that humans stop reading. The hours that used to be consumed by the reading get redeployed into the deciding, the talking, the writing, and the part of the job that no model has any business doing.

I would summarize the line we draw like this:

That split is the operating model we run on 39 accounts at RMC, and it is the operating model actcenter is built to deliver to any agency that wants out of spreadsheet hell without giving up the steering wheel.

The takeaway

The analyst day is the bottleneck — not because analysts are slow, but because the reading work was never structured. Move the reading into structured data and a model that can read it, and you do not just save time. You raise the floor on decision quality, because the analysis stops being a person and starts being a process.

Buy back the analyst day.

Two strategists, 30 days, no credit card. We wire the scripts to your MCC, route the data through Sheets, and your briefcase is ready by 8am tomorrow. You still hold the steering wheel.

Request a pilot → See pricing