---
title: "Why Claude Misses Waste on Big Ad Accounts—and How to Narrow the Ask"
description: "Big marketing MCP results can overwhelm Claude’s context. Filter, aggregate and rank in the tool call, then drill down into numbers you can check."
url: "https://groas.com/post/why-marketing-mcp-answers-go-wrong-on-big-accounts-context"
image: "https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/876601ba-0e71-4721-94c9-a8db8dc9754f.png"
published: "2026-11-20T14:39:18.214Z"
modified: "2026-11-20T14:39:18.281Z"
---

November 20, 2026 · 10 min read

# Why Claude Misses Waste on Big Ad Accounts—and How to Narrow the Ask

[](https://groas.com/author/alexander-perelman)

[Alexander Perelman](https://groas.com/author/alexander-perelman)Personally Tested

Head Of Product @ groas

[LinkedIn](https://www.linkedin.com/in/alexander-433793253/)

![A tiny clerk at a miniature desk buried under an endless printout proudly holds up one page with a red circled total, as if he had read it all.](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/876601ba-0e71-4721-94c9-a8db8dc9754f.png)

In this article

1. [The context window is the working surface, not the account](#the-context-window-is-the-working-surface-not-the-account)
2. [A polished answer can hide an incomplete report](#a-polished-answer-can-hide-an-incomplete-report)
3. [Why the full report may never reach Claude](#why-the-full-report-may-never-reach-claude)
4. [Five prompting patterns that shrink the work before it reaches chat](#five-prompting-patterns-that-shrink-the-work-before-it-reaches-chat)
5. [The three-line check before I act](#the-three-line-check-before-i-act)
6. [Apply the model to a six-location CPA spike](#apply-the-model-to-a-six-location-cpa-spike)

An 11,000-row search-term report is not a better brief for Claude. On a large ad account, it can be more data than the assistant can reason over well, wrapped in an answer that looks complete.

Say I’m reviewing a Google Ads account spending $20k a month. I ask Claude to pull every search term from the last 30 days and flag the waste. Seconds later: a clean table and a confident total. Neither tells me whether it actually analyzed the whole report.

The fix starts before the answer: **filter, aggregate and rank inside the tool call, then drill in.** Asking for everything and sorting it out in chat puts the hard work in the wrong place.

## The context window is the working surface, not the account

I think of the context window as the desk Claude gets to work on. Questions, loaded tool definitions and returned ad data all take up space. On a small account, that may be fine. On a big one, the desk can fill before the useful analysis starts.

Here is the running example: a Google Ads account with 40 campaigns, 800 ad groups and 11,000 search terms over the last 30 days. Through [groas as my MCP connector](https://groas.com/mcp), I ask Claude to “pull all search terms and find waste.” Depending on the fields returned, that result can run to hundreds of thousands of tokens.

The report is not the only thing competing for room. One published example describes three MCP servers consuming [143k of a 200k-token window before the first question](https://www.runagents.co/blog/best-mcp-servers-for-marketing-teams). A connection to the account does not give Claude unlimited working space.

The desk analogy has a limit: **fitting the data is not the same as analyzing it correctly.** A report can be incomplete, and a complete report can still be calculated badly. I need to check both.

## A polished answer can hide an incomplete report

An overloaded answer often looks finished. The warning signs are in what it leaves out:

- It lists 50 search terms when the account contains thousands, without labeling them as a subset.
- It names four campaigns but misses the biggest spender.
- It gives a total that I cannot reconcile with the account.

The last one is especially dangerous. A tidy dollar figure feels more authoritative than a partial list, even when both come from the same incomplete view.

In one published audit, an assistant reported $1,840 in waste, or 61% of budget, after counting the same cost twice for terms appearing in two ad groups. Checking the rows brought the figure to [$640, or 21%](https://advizify.com/blog/google-ads-audit-ai).

That example is an accounting failure, not proof of a context-limit failure. The distinction matters. Narrowing the result helps Claude inspect it, but I still need the grouping and calculation to be right.

**A confident total is a claim to verify, not evidence that every row was read.** Before changing a bid, I want to know what went into it.

## Why the full report may never reach Claude

The size of the account and the size of the tool response are different problems. The MCP specification itself sets no size limit on tool results, and its built-in pagination covers listing tools, not the data returned by an individual tool call. A tool therefore needs its own way to bound or page a large report.

If it returns too much, the application around the model may intervene. Claude Code, for example, has been documented as warning past [10,000 tokens of MCP output and capping it at 25,000 by default](https://www.rightshold.com/answers/how-to-cap-mcp-tool-output-size-to-prevent-context-flooding), saving longer output to a file rather than putting it all in chat.

That protects the window. It does not mean the assistant has analyzed the saved file.

Reported tests describe oversized results being truncated, failing transport, or reaching the model as a partial list. In one test involving 24,000 invoices, a cap passed only 10 rows to the model with no way to fetch the rest. One model identified the sample; another [multiplied it up to estimate a total](https://superglue.ai/blog/mcp-tool-result-limits/).

This is where a plausible answer becomes misleading: the assistant treats the visible slice as the whole account.

**My first defense is a bounded result, not a more forceful request to “analyze everything.”** The next five patterns put that into practice.

## Five prompting patterns that shrink the work before it reaches chat

Each pattern changes the report Claude receives, not just the wording of its answer. That distinction is the point. Asking Claude to display 20 rows after retrieving 11,000 does not solve the same problem as returning 20 rows from the tool.

**I specify the reduction in the tool call, then check that the returned result reflects it.** A prompt is an instruction, not proof that the tool performed the operation.

### 1. Filter and group before returning rows

For the $20k account, I do not start with every search term. I start with a narrower question: which terms spent more than $25 without a conversion in the last 30 days?

I also specify the grouping. A search term can appear in more than one ad group. If I want a term-level review, I need spend and conversions summed by term before applying the threshold.

My prompt:

> For the last 30 days, group search terms by term and sum spend, clicks and conversions. Return only terms with summed spend over $25 and zero summed conversions. Apply the grouping and filters in the tool call.

This follows the principle of [filtering, projecting and aggregating before listing](https://superglue.ai/blog/mcp-tool-result-limits/). The irrelevant rows stay out of the working context rather than arriving with instructions to ignore them.

**Define the date range, grouping and threshold before asking for the list.** Otherwise, “find waste” leaves too much of the report’s shape unspecified.

### 2. Build blended ratios from summed totals

Filtering controls volume. Aggregation also controls the math.

Suppose one campaign spends $100 to make $800 in revenue: 8.0x ROAS. Another spends $900 to make $1,800: 2.0x. Averaging those ratios gives 5.0x. Dividing total revenue by total spend gives [$2,600 / $1,000 = 2.6x](https://www.datanable.com/blog/why-ai-gets-marketing-numbers-wrong).

The second calculation answers the account-level question. The first gives both campaigns equal weight despite their different spend.

For the running account, my prompt is:

> Return total spend and total revenue for the last 30 days, computed from the full matching dataset. Compute account ROAS as total revenue divided by total spend. Show both totals and the formula.

That is the useful part of [defining the metric and formula before computing it](https://front.count.co/ai/prompts-for-data-analysis). I am not handing Claude 40 ratios and hoping it chooses the right way to combine them.

**Compute blended ratios from summed totals, never by averaging the component ratios.** A smaller answer is useful; a correctly defined answer is essential.

![A tiny desk buried under search-term reports, with one highlighted sheet pulled out](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/4acd36f6-6f1a-4731-ada7-ed7afc74e7ec.png)

### 3. Rank the result and cap it explicitly

A filtered list can still be too long. Next, I give it an order and a limit.

For an initial waste review, I want the highest-spend candidates first, not whichever rows happen to arrive first:

> For the last 30 days, group search terms by term and sum spend and conversions. Keep terms with spend over $25 and zero conversions. Sort by spend descending and return the top 20 only. Apply the limit in the tool call and state whether more matching terms exist.

That turns an open-ended request into a bounded review. It also makes the omission explicit: this is the top 20, not every matching term and not total account waste.

Good pagination makes the boundary visible through [a limit, a cursor and an indication that more results exist](https://dev.to/jeff_pdc/when-an-mcp-tool-returns-too-much-pagination-and-truncation-1458). When those are available, I want Claude to report them rather than quietly presenting page one as the whole result.

**A ranked subset is useful when it admits it is a subset.** I use it to choose where to investigate, not to declare the audit finished.

### 4. Drill down through the account structure

The running account has 40 campaigns, 800 ad groups and 11,000 search terms. I do not need all three levels at once to decide where to look first.

I work down the structure:

1. Return campaign-level totals for the last 30 days: spend, conversions and CPA.
2. Select the two campaigns with the worst CPA for closer review.
3. Return ad-group totals inside those campaigns only.
4. Inspect search terms inside the three worst ad groups.

Each answer narrows the next question. Instead of carrying the entire account through the conversation, I carry the decision about where to look.

Paging through everything is not automatically a practical alternative. At 10 rows per page, 24,000 rows require [2,400 calls](https://superglue.ai/blog/mcp-tool-result-limits/). A technically pageable report can still be the wrong workflow.

This route also has a boundary: inspecting selected campaigns does not establish what is happening everywhere else. **I label the scope of the finding as carefully as the finding itself.** “These campaigns need review” is not “the rest of the account is clean.”

### 5. Keep arithmetic in the tool or code

The second pattern defines the right formula. This one specifies where the calculation happens.

Even a correctly defined total is unreliable if Claude reconstructs it from a long list in chat. I want the tool to return the summed values, or the data processed in code, rather than asking the model to add rows in prose.

In a reported 12.4MB test, Claude, ChatGPT and Cursor got the right checksums only when they [saved the result to a file and queried it in code](https://superglue.ai/blog/mcp-tool-result-limits/) instead of reading all the rows in chat.

For the campaign view, my prompt is:

> Return one row per campaign for the last 30 days, with summed spend and summed conversions. Compute CPA as spend divided by conversions. Use the full matching dataset for the sums, not just the displayed rows.

That final sentence separates an account calculation from a page calculation. Twenty displayed rows may be enough for inspection; they are not automatically enough for an account total.

**I want quoted totals to trace to a tool result or a reproducible calculation.** A polished sentence is not a calculator receipt.

## The three-line check before I act

Before acting on a big-account answer, I ask Claude to provide three things:

1. **Coverage:** the date range, matching row count and any limit or truncation.
2. **Calculation:** the exact formula and where it was computed.
3. **Evidence:** the aggregate rows behind the total, plus the detail needed to inspect a flagged item.

I do not ask it to paste the entire report back into chat. That would recreate the problem. I want enough information to verify the scope and calculation, then drill into the relevant rows.

When I use [groas for marketing MCP workflows](https://groas.com/post/best-mcp-servers-for-marketing-in-2026-tested-on-real-ad), I keep that order in the prompt: establish coverage, define the metric, compute from totals. If the answer cannot show those basics, I check Google Ads before touching a bid.

The same sequence applies outside search terms. For 5,000 products, I would first make sure [ChatGPT is reading real ad data rather than guessing](https://groas.com/post/why-your-ai-assistant-guesses-about-your-ads-and-how-to-make). Then I would request summed spend and revenue by product for the last 30 days, minimum $50 spend, sorted by loss descending, top 25 only. After that, I would drill into the top three.

**Connecting the data is step one. Keeping the question narrow is step two.** Neither replaces the other.

## Apply the model to a six-location CPA spike

Now take a different case: a local services advertiser with six locations wants a full Q3 review across Google and Meta, every campaign, every keyword, plus Reddit mentions to explain a CPA spike.

That request fails the desk test before the first result arrives. It mixes account totals, detailed performance rows and unstructured text. I would split it into four passes:

1. **Locate the change.** Return summed spend, leads and cost per lead by advertising channel for Q3, one line per channel.
2. **Narrow the campaign set.** Return the top 10 campaigns by spend increase versus Q2, with the increase computed in the tool.
3. **Inspect search detail.** Within the two worst Google campaigns identified in the review, return search terms spending over $50 with zero booked jobs. Keep the result bounded and state whether more matches exist.
4. **Examine customer language separately.** Return Reddit quotes about price and scheduling from the last 90 days, capped at 20.

The first pass establishes the totals. The next two narrow the performance investigation. The last adds customer language without burying the account data under a pile of quotes. Those quotes may help interpret the spike; they do not establish its cause on their own.

Each answer now has a job, a scope and a boundary I can inspect. I can follow a number back to its aggregate instead of trusting a summary of an oversized export.

Small accounts grow. Date ranges widen. The habit of asking for everything survives until the month it breaks.

My rule is simple: **if my prompt has no date range, no threshold and no limit, I rewrite it before sending it.** Filter in the call, sum before listing, rank with a number and drill in steps. Give Claude a question it can answer from the data it actually receives, not a report it has to pretend it read.

## Frequently Asked Questions

### Why does Claude struggle with large ad account reports like an 11,000-row search term report?

On a large ad account, the full report can run to hundreds of thousands of tokens, and questions, tool definitions and returned data all compete for space in the context window. The desk can fill before the useful analysis starts, so fitting the data is not the same as analyzing it correctly.

### How can I tell if Claude's answer about ad waste is based on incomplete data?

Warning signs include a short list of search terms when the account contains thousands, campaigns mentioned while the biggest spender is missing, and a total that cannot be reconciled with the account. A confident dollar total is a claim to verify, not evidence that every row was read.

### Does Claude always receive the full result of an MCP tool call?

No. The MCP specification sets no size limit on tool results and its built-in pagination does not cover individual tool call data, so each tool must bound or page its own output. Applications like Claude Code have been documented as warning past 10,000 tokens of MCP output and capping it at 25,000 by default, so the model may only see a partial list or a truncated slice.

### How should I prompt Claude to find wasted spend without it pulling every search term?

Define the date range, grouping and threshold before asking for the list. For example, ask for search terms grouped by term with spend, clicks and conversions summed, returning only terms with summed spend over $25 and zero conversions for the last 30 days, with the grouping and filters applied in the tool call. Then check that the returned result actually reflects the requested reduction.

### Why is averaging campaign ROAS values wrong when calculating account-level ROAS?

Averaging gives each campaign equal weight despite different spend. For example, one campaign spending $100 to make $800 is 8.0x ROAS and another spending $900 to make $1,800 is 2.0x; averaging gives 5.0x, while dividing total revenue by total spend gives $2,600 divided by $1,000, which is 2.6x. The blended calculation from summed totals answers the account-level question.

### What is the best way to review waste on an account with 40 campaigns and 11,000 search terms?

Narrow the work in steps: return campaign-level totals for spend, conversions and CPA, pick the two worst campaigns, then review ad groups inside them and search terms inside the three worst ad groups. For lists, sort by spend descending, return a top 20 only, and state whether more matching terms exist. Label the scope of each finding, since inspecting selected campaigns does not establish that the rest of the account is clean.

### What should I check before acting on Claude's waste figures from a big ad account?

Ask Claude for three things before acting: coverage, meaning the date range, matching row count and any limit or truncation; calculation, meaning the exact formula and where it was computed; and evidence, meaning the aggregate rows behind the total plus detail to inspect a flagged item. Quoted totals should trace to a tool result or a reproducible calculation, and if the answer cannot show those basics, check Google Ads before touching a bid.

## Run Your Marketing On groas

Plug ChatGPT, Claude or any AI agent into groas and let it do the work of a whole marketing team: $19 a month flat with every feature included, and autopilot whenever you’re ready.

[Start For $19](https://groas.com/pricing)

[Or have groas run it for you](https://groas.com/done-for-you)

## Structured data

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://graph.groas.com/entity/groas","name":"groas","url":"https://groas.com/","description":"groas is the marketing infrastructure for your own AI: it connects your AI to all your marketing and gives it proprietary marketing data collected 24/7. Plug Claude, ChatGPT or any AI agent into groas and run all your marketing from it, for a tiny fraction of the usual cost, with the groas app there whenever you want to see more. It is the layer ChatGPT, Claude and any AI assistant run a business’s online marketing on. Through groas an assistant can do anything you could do natively in Google Ads, Bing Ads, Meta Ads and ChatGPT ads, on your website and in AI search, and, where the owner switches it on, can run connected campaigns by itself, so AI agents do the work a whole marketing team used to. groas also gives those agents large-scale proprietary marketing data it collects around the clock, on any schedule the owner chooses: every ad competitors run on Google and Meta, their landing pages and full funnels with the offers and prices in them, Reddit customer research, AI search answers from ChatGPT, Google AI Overviews, Gemini, Claude and Perplexity, and SEO data from Semrush and Ahrefs, backlinks included ($19 a month, flat, with everything included, for unlimited teammates on one shared workspace and unlimited ad accounts and websites, delivered as an MCP server). Separately, its done-for-you service has senior people and proprietary quantitative systems run paid ads, SEO and AI search for businesses, and white label for agencies, from $5,000 a month.","logo":{"@type":"ImageObject","url":"https://groas.com/icon-512.png","width":512,"height":512},"sameAs":["https://www.linkedin.com/company/groas/"]},{"@type":"WebSite","@id":"https://groas.com/#website","name":"groas","url":"https://groas.com/","inLanguage":"en-US","publisher":{"@id":"https://graph.groas.com/entity/groas"}},{"@type":"WebPage","@id":"https://groas.com/post/why-marketing-mcp-answers-go-wrong-on-big-accounts-context#webpage","url":"https://groas.com/post/why-marketing-mcp-answers-go-wrong-on-big-accounts-context","name":"Why Claude Misses Waste on Big Ad Accounts—and How to Narrow the Ask","description":"Big marketing MCP results can overwhelm Claude’s context. Filter, aggregate and rank in the tool call, then drill down into numbers you can check.","inLanguage":"en-US","isPartOf":{"@id":"https://groas.com/#website"},"about":{"@id":"https://graph.groas.com/entity/groas"},"primaryImageOfPage":{"@type":"ImageObject","url":"https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/876601ba-0e71-4721-94c9-a8db8dc9754f.png"},"breadcrumb":{"@id":"https://groas.com/post/why-marketing-mcp-answers-go-wrong-on-big-accounts-context#breadcrumb"},"dateModified":"2026-11-20T14:39:18.281Z","mainEntity":{"@id":"https://groas.com/post/why-marketing-mcp-answers-go-wrong-on-big-accounts-context#article"}},{"@type":"BreadcrumbList","@id":"https://groas.com/post/why-marketing-mcp-answers-go-wrong-on-big-accounts-context#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://groas.com/"},{"@type":"ListItem","position":2,"name":"Blog","item":"https://groas.com/blog"},{"@type":"ListItem","position":3,"name":"Why Claude Misses Waste on Big Ad Accounts—and How to Narrow the Ask","item":"https://groas.com/post/why-marketing-mcp-answers-go-wrong-on-big-accounts-context"}]},{"@type":"BlogPosting","@id":"https://groas.com/post/why-marketing-mcp-answers-go-wrong-on-big-accounts-context#article","url":"https://groas.com/post/why-marketing-mcp-answers-go-wrong-on-big-accounts-context","mainEntityOfPage":{"@id":"https://groas.com/post/why-marketing-mcp-answers-go-wrong-on-big-accounts-context#webpage"},"isPartOf":{"@id":"https://groas.com/blog#blog"},"headline":"Why Claude Misses Waste on Big Ad Accounts—and How to Narrow the Ask","description":"Big marketing MCP results can overwhelm Claude’s context. Filter, aggregate and rank in the tool call, then drill down into numbers you can check.","datePublished":"2026-11-20T14:39:18.214Z","dateModified":"2026-11-20T14:39:18.281Z","image":{"@type":"ImageObject","url":"https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/876601ba-0e71-4721-94c9-a8db8dc9754f.png"},"author":{"@id":"https://groas.com/author/alexander-perelman#person"},"publisher":{"@id":"https://graph.groas.com/entity/groas"},"keywords":"claude, account, tool, rows, spend, total, waste, before","wordCount":2192,"inLanguage":"en-US"},{"@type":"Person","@id":"https://groas.com/author/alexander-perelman#person","name":"Alexander Perelman","jobTitle":"Head Of Product @ groas","description":"My background includes Goldman Sachs and Stanford Computer Science.","image":"https://groas.com/media/blog/4f17cac3a81acc224cc1b2cfab8f94bbc295a3089d13e8421125771ef8d4064f.jpg","worksFor":{"@id":"https://graph.groas.com/entity/groas"},"sameAs":["https://www.linkedin.com/in/alexander-433793253/"]},{"@type":"FAQPage","@id":"https://groas.com/post/why-marketing-mcp-answers-go-wrong-on-big-accounts-context#faq","mainEntity":[{"@type":"Question","name":"Why does Claude struggle with large ad account reports like an 11,000-row search term report?","acceptedAnswer":{"@type":"Answer","text":"On a large ad account, the full report can run to hundreds of thousands of tokens, and questions, tool definitions and returned data all compete for space in the context window. The desk can fill before the useful analysis starts, so fitting the data is not the same as analyzing it correctly."}},{"@type":"Question","name":"How can I tell if Claude's answer about ad waste is based on incomplete data?","acceptedAnswer":{"@type":"Answer","text":"Warning signs include a short list of search terms when the account contains thousands, campaigns mentioned while the biggest spender is missing, and a total that cannot be reconciled with the account. A confident dollar total is a claim to verify, not evidence that every row was read."}},{"@type":"Question","name":"Does Claude always receive the full result of an MCP tool call?","acceptedAnswer":{"@type":"Answer","text":"No. The MCP specification sets no size limit on tool results and its built-in pagination does not cover individual tool call data, so each tool must bound or page its own output. Applications like Claude Code have been documented as warning past 10,000 tokens of MCP output and capping it at 25,000 by default, so the model may only see a partial list or a truncated slice."}},{"@type":"Question","name":"How should I prompt Claude to find wasted spend without it pulling every search term?","acceptedAnswer":{"@type":"Answer","text":"Define the date range, grouping and threshold before asking for the list. For example, ask for search terms grouped by term with spend, clicks and conversions summed, returning only terms with summed spend over $25 and zero conversions for the last 30 days, with the grouping and filters applied in the tool call. Then check that the returned result actually reflects the requested reduction."}},{"@type":"Question","name":"Why is averaging campaign ROAS values wrong when calculating account-level ROAS?","acceptedAnswer":{"@type":"Answer","text":"Averaging gives each campaign equal weight despite different spend. For example, one campaign spending $100 to make $800 is 8.0x ROAS and another spending $900 to make $1,800 is 2.0x; averaging gives 5.0x, while dividing total revenue by total spend gives $2,600 divided by $1,000, which is 2.6x. The blended calculation from summed totals answers the account-level question."}},{"@type":"Question","name":"What is the best way to review waste on an account with 40 campaigns and 11,000 search terms?","acceptedAnswer":{"@type":"Answer","text":"Narrow the work in steps: return campaign-level totals for spend, conversions and CPA, pick the two worst campaigns, then review ad groups inside them and search terms inside the three worst ad groups. For lists, sort by spend descending, return a top 20 only, and state whether more matching terms exist. Label the scope of each finding, since inspecting selected campaigns does not establish that the rest of the account is clean."}},{"@type":"Question","name":"What should I check before acting on Claude's waste figures from a big ad account?","acceptedAnswer":{"@type":"Answer","text":"Ask Claude for three things before acting: coverage, meaning the date range, matching row count and any limit or truncation; calculation, meaning the exact formula and where it was computed; and evidence, meaning the aggregate rows behind the total plus detail to inspect a flagged item. Quoted totals should trace to a tool result or a reproducible calculation, and if the answer cannot show those basics, check Google Ads before touching a bid."}}]}]}
```
