Most teams treat content like a line item. That is the first mistake.

The second is treating it like a lottery ticket. Write a hundred pieces, hope five or six hit, and declare the program a success because the spreadsheet still has green cells in it. That worked, sort of, when content was expensive and slow to produce. It doesn't work now.

AI has made it trivial to produce volume. It has not made it any easier to produce content that works. If anything, the flood of generated copy has raised the bar for what counts as effective and made the cost of guessing wrong much higher.

So the question is no longer how much content you can ship. It's whether you can predict, before you publish, which pieces will actually do the job.

Several of the ideas below come from a conversation between Colleen Jones, founder of Content Science and author of The Content Advantage, and Jeff Coyle about how teams predict and measure content effectiveness. They hold up better now than when they were first said.

Investment Versus Cost

Jones sees a consistent divide between organizations that view content as an investment and those that view it as a cost. The first group has a defined budget, a defined return, and a defined way of measuring whether the two line up. The second group has invoices.

The tell is simple. Ask a content team what they spent last quarter. The investment group can break it down into everything that touches a published asset:

  • People: writers, editors, and strategists, whether in-house, freelance, or agency
  • Platform: the CMS and the rest of the publishing stack
  • Distribution: email, social, and every other channel that carries the content
  • Promotion: paid spend that puts the content in front of people

The cost group says something like, "We paid a freelancer $1,000 an article and published six." That is not a budget. That is a receipt.

Coyle draws the same line from the other side. Leaders who can name the true cost of content, with all its components, tend to run programs that can show what they return.

The gap matters because everything downstream, from measurement to prediction to optimization, depends on knowing what a piece of content actually costs to produce. You cannot calculate return on an investment you have not bothered to define. The CMO Council's work on content ROI makes the point from the budget side. Once content takes a real share of marketing spend, not knowing its cost stops being a rounding error.

The 20/80 Problem

Jones describes a familiar pattern. When little thought goes into strategy, roughly 20% of the content produces 80% of the results. Sometimes it's 30/70. Sometimes it's worse. The shape doesn't change.

This is what happens when strategy is thin and data is thinner. Teams publish what they think will work, find out months later which pieces actually did, and repeat the cycle. The variance is enormous because the inputs are guesses.

Coyle puts numbers on the hit rate. Most teams land somewhere between 5% and 30%. Of every hundred articles they publish, between five and thirty are worth having published. The rest exist.

That hit rate changes what content really costs. Say you pay $500 an article and one in ten does its job. Each piece that works cost you $5,000. Raise the hit rate to 40% and the same $500 article costs $1,250 per success. The price per article never changed. What you got for it did.

Before you count hits, decide where to count them. Judging each page on its own misses how topics behave. A group of pages on one subject tends to rise together, so a supporting page can get little direct traffic while lifting the whole cluster. Evaluate at the page level and at the cluster level, and count a support page that does its supporting job as a hit.

The point of a data-driven approach isn't to eliminate variance. It's to reduce it. If you can move the hit rate from 20% to 50%, you have not changed the cost of producing content. You have changed what that content is worth.

Prediction Is Variance Reduction

Prediction in content is not fortune-telling. It is the boring work of collecting enough signal that the next piece is less of a guess than the last one. Which topics resonate. Where authority already exists. What the audience is actually searching for.

This is where most programs fall apart. They have analytics. They do not have a system that feeds those analytics back into decisions about what to write. The data sits in dashboards. The writers sit in a different room.

The cheapest place to close that gap is the outline. Before a writer spends a week on a draft, check the plan against what thorough coverage of the topic requires: the subtopics, the questions readers ask, the related concepts that signal authority. The gaps you find there are a forecast of where the finished piece will lose ground. Fixing them at the outline stage takes an hour. Fixing them after the page underperforms takes a rewrite.

Content intelligence exists to put that signal in front of the strategist before the brief is written, not after the piece underperforms. A good strategy depends on being able to act on what the data already knows.

Intent Mismatch

Here is the failure mode that quietly burns the most budget: the searcher wanted one thing, the page delivered another, and nobody noticed because the page still ranked.

Coyle calls it intent mismatch, the gap between what brought someone to the page and what the page actually does. It happens because search behavior is messy and writers work from briefs that focus on keywords rather than the question behind the keyword. A page that defines what a content brief is starts ranking for "content brief templates." It has no templates. The traffic arrives and leaves.

You can see it in the analytics if you know what to look for. High traffic, high bounce, low scroll depth, no downstream action. The page is doing its job by one definition and failing by every other.

Fixing intent mismatch is unglamorous work. It's auditing what you already have, identifying the misses, and either rewriting the page to match the intent or building a new page that does. It is also one of the highest-leverage things a content team can do, because the traffic is already arriving.

The Vanity Metric Problem

Pageviews are not effectiveness. Neither is time on page, social shares, or any of the other numbers that look good in a quarterly slide and tell you almost nothing about whether the content moved the business.

Jones recommends a mix of three kinds of data working together: behavioral analytics (what people did), voice-of-customer data (why they did it), and business metrics (whether any of it mattered). One without the others gives you a misleading picture.

That mix now needs a fourth layer. A growing share of discovery happens inside AI answers, where your content is summarized, cited, or left out before anyone clicks. Behavioral analytics can't see any of that. A page can shape how an answer engine describes your category and never register a visit. Tracking whether your content is mentioned and cited in AI answers closes the AEO gap in content measurement, the blind spot the other three sources share.

The over-reliance on scale alone is what makes AI-generated content so dangerous. It is now possible to produce a thousand articles in a week and watch your pageviews climb while every meaningful business metric flatlines. The dashboard will tell you it's working. It isn't.

The Two Measurement Extremes

Jones finds that most teams sit at one of two extremes.

The first ignores measurement entirely. Maybe once a quarter someone pulls a report. Maybe it's outsourced to analytics people who don't know the content. The team is flying blind and doesn't know it.

The second drowns in measurement. Every meeting is a data argument. Every decision waits for one more dashboard. The team has convinced itself the numbers will decide what to do, which is a polite way of saying nobody wants to.

Neither extreme produces good content. The first because it has no feedback loop. The second because it has nothing but feedback loops and no judgment.

The fix is not more data. It is the right data, used to inform decisions a person still has to make. Creating the content is still a creative act. Measurement makes it less of a gamble, not less of a craft.

What This Looks Like in Practice

Predicting and measuring content effectiveness comes down to a few unglamorous habits.

Define what content actually costs, all of it, not just the freelancer invoice. Divide that cost by your hit rate to see what each effective piece really costs. Judge hits at the page level and the cluster level. Build a measurement framework that combines behavioral data, customer feedback, business outcomes, and AI visibility. Check coverage at the outline stage, before anyone writes. Audit existing content for intent mismatch before producing anything new. Treat prediction as variance reduction, not prophecy.

And stop counting pageviews like they're proof of anything.

When it's time to turn those numbers into a case leadership will fund, see how to measure and communicate content ROI.

Siteimprove.ai puts content intelligence and SEO data in one platform, so deciding what to write next and checking how existing content performs don't happen in separate tools.

The teams that win the next few years won't be the ones producing the most content. They'll be the ones who knew which pieces to write before they wrote them.