Case study · Social research design
A page had been running for over a decade. During long stretches when nothing new was posted, people kept commenting — and not only the regulars. New people kept arriving on old material. That is not how content is supposed to behave. This is how we turned an anecdote into something testable.
Everything in social media marketing assumes content is perishable. You post, you get a spike, the spike decays, you post again. Every planning tool, every content calendar and every agency retainer is built on that assumption.
The page in question is an archive of historical photographs. Its behavior did not match. Old posts kept collecting comments months and years later, from people who had never engaged with the page before, during periods when nothing new was being published at all.
An owner noticing a pattern on their own page is the weakest possible evidence. They see what they hope to see, they remember the good weeks, and they have no comparison group. The observation was interesting precisely because it was inconvenient — so the first job was to design something that could prove it wrong.
What is actually being claimed
This diagram is the hypothesis, drawn. It is not measured data. Establishing whether the solid line is real is the purpose of the study described below.
If the claim holds, it changes the arithmetic of a whole category of business — museums, historical societies, publishers, archives, heritage tourism, anyone whose material is old on purpose. Their content budget is being planned on a decay curve that may not apply to them.
The study has two halves. One looks backward at data that already exists. The other runs forward and can return a null result.
Ten years of comment timestamps on a single page is a real dataset. Measured against the publication date of each post, it produces an actual decay function rather than an impression of one. It also answers the sharper question: are the late comments coming from the same people, or from new arrivals? Those two findings mean completely different things.
The retrospective data can only show that the pattern exists. It cannot show why. So a second component tests one specific explanation — that provenance is what makes archival material durable.
One variable, three treatments
Design note: the arms differ only in how much sourcing and context accompany the image. Assignment, timing and posting cadence are held constant so that any difference in the decay curve is attributable to the treatment.
| Component | Question it answers | What a negative result would look like |
|---|---|---|
| Retrospective decay curve | Does archival content decay more slowly than the conventional pattern? | It decays the same. The observation was selective memory. |
| New versus returning commenters | Is the tail new audience, or the same regulars? | Same regulars. It is a community effect, not a content effect. |
| Provenance experiment | Does sourcing and context cause the durability? | No difference between arms. Something else explains it. |
| Economic model | What is an archive worth if the durability is real? | Nothing, if the first three fail. |
Alongside the page data, we built a picture of the wider category — the independent accounts, groups and pages already publishing in this subject area. Several hundred of them. That inventory does two separate jobs, and keeping them separate matters.
| As research | As a campaign input |
|---|---|
| A population to observe: how large the community is, how it clusters, what kinds of material circulate | Evidence that an audience exists at scale, which is what a commercial plan has to demonstrate |
| Observed only. Nothing is posted to these channels and they are not participants in the experiment | Used to size the opportunity, not to claim ownership of it |
The same landscape work, in an adjacent project, produced the audience figure that turned a publisher’s “maybe” into a six-volume commission. That story is told separately — see the publishing case study.
The commercial page and the research are deliberately kept apart. The page used for publishing and promotion is a separate property from the pages being studied, the research is not run for profit, and the third-party channels in the inventory are observed and never posted to. That separation was written into the protocol at the start rather than defended afterward, because a study that doubles as a marketing campaign is neither.
The reason to run this at all is that it rewrites how the asset gets valued.
Perishable content, or a stock of assets
The right-hand framing treats each archival item as producing a small recurring return - an engagement annuity - so the archive is valued as a stock rather than as this month's output. The measurement has to come first.
There is a practical version of this for anyone sitting on old material. If a photograph posted in 2016 is still bringing new people to a page in 2026, then digitizing the next thousand photographs is a capital project with a return, not a marketing expense with a shelf life. Those two things get budgeted very differently and approved by different people.
An earlier version of this plan carried follower-growth projections and an exit valuation. They were built the way most social media plans are built — a growth rate assumed, compounded forward, multiplied by a number found in an article.
They were removed. Not because they were pessimistic or optimistic, but because nothing underneath them was measured. A projection that cannot be checked is not a forecast; it is a decoration. The study design above is what took their place, and it is a weaker-looking document that is worth considerably more.
The protocol is designed and the retrospective dataset is in hand. Results are not in, and this page will not claim any. What is being shown here is the method — how a pattern somebody noticed gets converted into something that can be measured, argued with, and published.
| Stage | Status |
|---|---|
| Observation recorded and stated as a falsifiable claim | Done |
| Retrospective dataset assembled | In hand |
| Landscape inventory of the wider category | Done |
| Ethical separation of research and commercial properties | Written into the protocol |
| Decay function measured | Not yet |
| Provenance experiment run | Not yet |
| Economic model calibrated against real figures | Depends on the two above |
Most social media strategy is written from received wisdom: post this often, on these channels, in this format. It is rarely tested against the business it is written for, and almost never against material that behaves differently from the norm.
What is shown here is the alternative. Notice the thing that does not fit. State it in a way that can be proven wrong. Build a design that would expose it if it is. Draw the ethical lines in writing, at the start. And decline to publish numbers you have not yet earned.
Almost every organization with an archive — a museum, a newspaper, a historical society, a family firm with eighty years of photographs — is sitting on the same untested question. Nobody has measured whether their old material behaves like everyone else’s new material. It is a cheap thing to find out and an expensive thing to assume.
Marketing Analytics Consultants
We design the study, run the numbers and tell you what they support — including when the answer is that the idea does not hold up.