The 70/30 rule: why great performance marketing is mostly business context

Great performance marketing is 70 percent business context and 30 percent ad platform knowledge. Every AI tool on the market has solved the 30 percent. The other 70 percent remains largely outside every system, and that gap is why AI recommendations so often miss the mark.

What the 30 percent looks like

The 30 percent is the part the tools have genuinely nailed. Bid strategy logic. Audience segmentation mechanics. Creative format specs. Quality score factors. Smart bidding signals. Attribution windows.

This is the knowledge that lives in documentation, in certification courses, in platform changelogs. It is learnable, codifiable, and stable enough that a model trained on it will give you a competent answer almost every time.

Ask any major AI tool to explain broad match modifier behavior, or to audit a Google Ads account structure against best practices, and you will get a solid answer. That problem is solved.

What the 70 percent actually contains

The 70 percent is harder to name because it does not live in any single place. It is the accumulated reasoning behind every decision in the account.

Here is a concrete inventory:

• Why the account is structured the way it is. There was a reason someone consolidated those campaigns in Q3. It is not in the account. It is in the head of the person who ran the meeting.
• What you tested and killed. You paused that ad set in March because the creative failed, not because the targeting was wrong. That distinction matters enormously if someone tries to run the same test in July.
• What leadership decided on Monday. The budget shifted. The KPI changed. A competitor launched. The strategy call happened and the account has not caught up yet.
• What Legal cleared and what it did not. A claim that worked in copy last quarter may have been flagged since. No system knows this unless someone typed it somewhere.
• Why certain clients or verticals are off-limits. Exclusions that exist for relationship reasons, not performance reasons. They are invisible to any tool reading the ad account.
• The seasonal patterns your team has observed. Not the platform's generic seasonality signals, but the specific knowledge that your category spikes two weeks before the platform's data suggests it does.

None of this is in the ad account. None of it is in a dashboard. It is created continuously, in meetings and decisions and Slack threads, and it gets captured occasionally, manually, and partially.

Why it never makes it into the system

The capture problem is not laziness. It is structural.

Context is generated faster than any team can document it. A Monday strategy call produces a dozen decisions. A client email thread shifts three priorities. A Slack exchange between two account managers resolves a question that will come up again in six months. By the time anyone thinks to write it down, the next week has already started.

Even when teams do document, the documentation goes stale immediately. A brief written in March reflects March's strategy. By June it is a liability, not an asset, because someone will trust it.

And even when documentation is current, it does not accumulate. A prompt window holds weeks of context, not years. The learnings from last peak season do not automatically inform this one. When your best account lead resigns, the reasoning behind the account structure leaves with them.

The result is a widening gap. The longer a team runs an account, the more context they accumulate in people's heads, and the further that context drifts from anything a system can see.

What happens when AI operates without the 70 percent

This is where the mismatch becomes expensive.

An AI tool reading only the ad account will recommend scaling a campaign the team decided to sunset on Monday. It will propose a creative angle built around a claim Legal killed last month. It will suggest running a test that failed in March, with full confidence, because it has no record of March.

These are not edge cases. They are the normal output of a system operating on 30 percent of the available information.

The practical consequence is that every AI recommendation requires a senior review against account reality. Someone has to hold the 70 percent in their head and filter the output against it. That review does not go away just because the tool got faster. It may actually get heavier, because faster tools produce more output to check.

This is a significant part of why AI has not reduced headcount per account at most agencies. The tool handles the platform mechanics. The human still handles the context.

Why more data does not solve it

The instinct is to connect more sources. If context is missing, add the CRM. Add the Slack integration. Add the meeting transcripts.

The problem is that raw data volume is not the same as usable context. Connect everything without prioritization and the agent weighs a throwaway Slack message the same as a CEO strategy decision. Fourteen recommendations appear, with no signal about which one matters this week.

Context without classification is noise. The 70 percent is not just a volume problem. It is a relevance, recency, and business impact problem. A competitor launch this week should outrank a brand guideline update from eight months ago. A budget decision made Monday should immediately supersede a brief from March. That kind of prioritization does not happen automatically when you add more integrations.

The compounding problem over time

There is a longer-term cost that is easy to miss.

Platform knowledge compounds slowly because platforms change incrementally. But business context compounds fast. Every test adds a data point. Every quarter adds a pattern. Every personnel change is a risk to continuity.

Teams that run accounts for years develop a kind of institutional intelligence that is genuinely hard to replicate. They know which creative angles have been exhausted. They know which audience segments overlap with a competitor's core customer. They know what the brand will and will not approve.

None of that intelligence is in the system. It is in the people. Which means it is fragile. It leaves when they leave. It degrades when they get stretched across more accounts. It is invisible to any tool that reads only the ad account.

The teams that are furthest ahead are not the ones with the best AI tools. They are the ones who have figured out how to make that accumulated intelligence durable.

What solving the 70 percent actually requires

It requires capture that is continuous, not manual. Context has to be read from the places it is actually created: meeting recordings, Slack, email, ad platforms, CRM, business data. Not uploaded occasionally. Read automatically, as it happens.

It requires currency. Every meeting, experiment, and decision has to update the context in real time. A budget shift on Monday has to be reflected on Monday, not when someone remembers to update a brief.

It requires accumulation. Tests, results, and decisions need to be versioned and retained across quarters and years, not just held in a prompt window.

And it requires prioritization. Context has to be classified by relevance, recency, and business impact before an agent ever sees it. The CEO's strategy call should outrank a throwaway Slack thread. The current campaign objective should outrank last quarter's.

Where Almond fits

Almond was built specifically around this problem. The Business Context Engine reads ad accounts, Slack, Notion, Sheets, email, meeting notes, and CRM, captures every test, result, and decision as it happens, keeps it current, and ranks it by relevance, recency, and business impact before any agent or workflow touches it.

The products built on top of it, Canvas for AI-powered visual dashboards, Workflow Builder for judgment-based Google and Meta automations, and open integrations via MCP, all draw on the same context layer. A workflow will not scale a campaign the team decided to sunset on Monday, because the engine already knows about Monday.

Almond 2.0 early access is open at almond.ai. Every member gets a detailed 1:1 Meta and Google ad account audit with the CEO, recorded. Waitlist is work email only.

The 30 percent problem is solved. The 70 percent is where the actual work is.