The content half-life model

Every article you publish is either an asset or an expense, and the difference is measurable. This is the arithmetic that tells you which one you just bought — and the rule that tells you when to refresh instead of publish.

Content marketing is sold on an unexamined premise: that publishing is cumulative. Write fifty articles this year and you will have fifty articles working for you. The library grows, the traffic grows, the compounding happens on its own.

It does not work like that, and most teams find out eighteen months in, when output has doubled and traffic has not. The library did grow. The traffic did not, because a large share of what was published was never an asset in the first place. It was an expense with a slow burn.

The useful question is not “how much should we publish?” It is “what is the half-life of the thing we are about to publish, and does that half-life justify the cost of making it?” That question has an actual answer, and you can compute it from data you already have.

Two populations, not one average

The mistake underneath most content reporting is treating a content library as a single population and looking at its average performance. Averages hide the only distinction that matters.

A content library is really two populations that behave in opposite ways:

Reported together, these two populations produce a flat or gently rising average that tells you nothing. A library that is 80% decaying and 20% compounding can post the same “average traffic per article” as one that is the reverse — while being worth a fraction as much, because in the first case almost all the value is already gone and in the second almost all of it is still ahead.

So the first move is not to measure harder. It is to split the population.

The model

For decaying content, traffic after publication follows a shape close to exponential decay. That gives us a half-life: the time it takes for a page’s monthly traffic to fall to half of its peak.

T(t) = P × (1/2)^(t / h)

  T(t) = traffic in month t after publication
  P    = peak monthly traffic (usually month 1 or 2)
  h    = half-life, in months
  t    = months since publication

Total lifetime traffic is the area under that curve. Integrating gives a clean and genuinely useful result:

L = P × h / ln(2)   ≈   P × h × 1.4427

  L = total lifetime visits the page will ever deliver

This single line does more work than most content dashboards. It says lifetime value is linear in the half-life. A page with a six-month half-life is worth exactly twice a page with a three-month half-life at the same peak — and the peak is the number everyone optimises for while the half-life is the number nobody measures.

Compounding content does not fit this shape at all. Its curve is better described as bounded growth — slow start, accelerating middle, plateau:

T(t) = C × (1 - e^(-t / τ))

  C = ceiling: the traffic the page settles at
  τ = time constant, in months (time to reach ~63% of the ceiling)

The practical consequence: compounding content has no meaningful lifetime total, because the integral does not converge while the page stays relevant. You cannot compare the two populations with one metric. You have to hold them apart, which is exactly what averaging refuses to do.

Deriving your own half-life

None of this is useful with borrowed numbers. Published “content decay” benchmarks are close to meaningless across sites, because half-life is a property of the intersection of your topic, your authority and your competitors’ refresh behaviour. Here is how to get your own figure. It takes about an hour.

1. Take a clean cohort

   Export every URL published in a single quarter at least fifteen months ago. Fifteen is the minimum: you need enough post-publication months to see the shape, and you want the cohort old enough that its trajectory has resolved.

2. Pull monthly organic traffic per URL

   Search Console, page-level, month by month from publication to now. Not sessions from analytics — organic clicks. You are measuring search demand capture, and mixing in referral and social spikes will corrupt the curve.

3. Normalise each page to its own peak

   For every URL, find its best month and divide every month by that value. Each page now runs from 1.0 down. This is what lets you compare a page that peaked at 40 visits with one that peaked at 4,000.

4. Classify, then split

   A page whose normalised traffic is still above 0.7 at month twelve is compounding. Below 0.4, it is decaying. Between the two, it is ambiguous — set those aside rather than forcing them, because a forced classification is where this analysis usually goes wrong.

5. Read the half-life off the decaying group

   For each decaying page, find the month at which normalised traffic first crosses 0.5. The median of those months is your half-life. Use the median, not the mean: this distribution has a long tail and the mean will flatter you.

The refresh rule

Once you have h, the most valuable decision in content operations stops being a matter of taste. Refreshing an existing page and publishing a new one compete for the same hours, and you can price both.

A refresh restores a decayed page toward its former peak. Its value is the difference between the traffic the page would have delivered untouched and what it delivers restored. For a page currently at fraction f of its peak, refreshed back to peak:

Value of refreshing:   V_r = P × (1 - f) × h / ln(2)
Value of publishing:   V_p = P_new × h / ln(2)

Refresh wins when:     P × (1 - f)  >  P_new × (C_r / C_p)

  f   = current traffic as a fraction of the page's peak
  C_r = cost to refresh      C_p = cost to publish new

The h / ln(2) term appears on both sides and cancels, which is the first genuinely surprising result of this model: the refresh-versus-publish decision does not depend on your half-life at all. It depends only on how far the page has fallen, how good the new page would be, and the relative cost of the two actions.

Since a refresh typically costs a fraction of a new commission, the threshold sits far lower than most teams assume. If refreshing costs a third of publishing, refreshing a page that has fallen to half its peak beats publishing a new page that would perform up to 1.5× as well as the old one did. That is a high bar for the new page, and it is usually not met.

A worked example

The numbers below are illustrative inputs, not benchmarks. Substitute your own from the method above.

Recoverable traffic from the refresh is 820 × (1 − 0.36) = 525 monthly clicks. Lifetime, that is 525 × 7 / ln(2) ≈ 5,300 clicks for £400, or about 7.5 pence per click.

For a new page to beat that, it would need to peak above 525 × (1400/400) = 1,838 monthly clicks — more than double the best month the existing page ever had. If you genuinely have that opportunity available, publish it. Most content calendars do not, which is why most content calendars are misallocated.

What this changes

• Report the two populations separately. A single “blog traffic” line is an average across opposite behaviours and will mislead you every quarter.
• Measure half-life once a year, not once. It moves as your authority and your competitors’ behaviour change.
• Price every commission against the refresh alternative before it enters the calendar, not after.
• Stop treating decaying content as failure. Event-anchored content can be entirely rational — it just has to be bought knowingly, at a price that matches a short half-life.
• Judge writers and agencies on the compounding share of what they produce, not on volume or on average traffic.

Where this model breaks

An honest model states its limits, so here are the four that matter.

It assumes traffic is the unit of value. It usually is not. A page delivering 60 visits a month from people about to buy can be worth more than one delivering 6,000 from people who will never buy. If you can attach revenue per visit by page, substitute that for traffic throughout — the algebra is unchanged and the conclusions get considerably sharper.

It assumes refreshes restore toward the original peak. Sometimes they exceed it, when the refresh genuinely improves the page against current intent. Sometimes they recover almost nothing, when the page lost position to a structurally stronger competitor rather than to ageing. Track your realised recovery rate and feed it back in as a coefficient.

It says nothing about whether the page should exist. A high half-life on a topic your buyers do not care about is an efficiently produced irrelevance. This model optimises allocation within a strategy; it does not supply one.

Questions