Original research
Original research, and why there is none here yet
This is where our original research will be published. There is none yet, and rather than dress that up we have written down the standard a study has to meet before it appears — a standard most published marketing benchmarks would not clear.
The state of this page
There is no research published here yet. When there is, it will be work we ran ourselves, with the sample, the method and the collection dates on the page.
We could fill this section tomorrow. The raw material is freely available, and the standard practice in this industry is to take a public dataset, recut it, give it a title with a year in it and call the result a study.
That is not research, and the version of this page that did it would be less useful to you than the empty one you are reading.
The reasoning
Why we would rather publish nothing than a benchmark we did not run
A benchmark is a claim about the world. Once published it gets quoted, then quoted without a link, then repeated by systems that summarise the web — and at that point it is effectively beyond correction. Publishing one carelessly does more damage than publishing nothing at all.
The tell is always the same. A study with a real method tells you how many things it looked at, where they came from, when they were measured and what it could not see. A study without one hands you a headline percentage and a chart. It is worth noticing how much of the marketing data you have read this year falls into the second group.
We are a new company, so here is the obvious thing said out loud: we do not yet have the volume of first-party data that would make some of these studies meaningful, and we are not going to disguise a small sample as an industry benchmark. Where a study is possible from genuinely public sources, the sample will be described precisely enough that somebody could argue with it.
The bar
What a study has to clear before it appears here
Written before any study exists, so it cannot quietly be adjusted later to fit a result we happened to like.
Every study will state
- The question it set out to answer, and why that question is worth asking now
- Where the data came from, named specifically, and whether it is public, first-party or licensed
- The sample: what was measured, how many of them, and how they were selected
- How it was collected, in enough detail that someone with the same access could repeat it
- The dates the data covers, on the page itself rather than only in the title
- What the study cannot tell you, and the ways it could be wrong
No study will
- Present a correlation as a cause because the headline reads better that way
- Report a percentage without the count it was calculated from
- Generalise from one sector or one country to "the industry"
- Use a sample chosen because it produced an interesting answer
- Recycle another organisation’s figures and present the result as ours
- Stay published unmarked once we know that part of it is wrong
Intended programme
The areas we intend to study, and what would make each worth doing
How AI answer engines choose their sources
Which pages get cited when an assistant answers a commercial question, and what those pages have in common structurally. Worth studying because almost everything written about it so far is inference drawn from a handful of screenshots.
Real-user performance by industry
Field performance varies enormously by sector, and public datasets make an honest comparison possible without touching a single client account. The value would be in the segmentation, not in yet another site-wide average.
How local businesses actually surface in map results
Observable, repeatable and badly served by the existing writing, most of which restates the same guidance year after year without testing whether it still holds.
What a marketing budget actually buys
A study we can only run properly once we have enough engagements of our own to describe honestly. Until then, saying so is more useful than publishing a model dressed up as evidence.
Website conversion patterns by buyer type
The gap between what agencies assert about conversion and what anyone can demonstrate is unusually wide here. Anything we published would need permission from every business whose data appeared in it, which is a slow constraint we accept.
In the meantime
Where the substantiated writing actually is
Writing with its sources attached
Questions
Fair questions about an empty research page
When will you publish your first study?
When one clears the standard set out on this page. We are not going to name a date, because a deadline is exactly the pressure that turns a thin dataset into a published benchmark.
The likeliest first candidates are the ones that can be run entirely from public data, because they do not depend on having accumulated years of first-party numbers first.
Why not summarise research other people have published?
We do, inside our writing, with a link to the original and the period it describes. What we will not do is repackage someone else’s figures under a new title and present the result as our research.
That practice is the reason so many marketing statistics have no traceable origin. Each retelling drops a little context until the number is being quoted with no method, no sample and no year attached to it.
Would you publish the underlying data?
Where we can, yes. If a study is built from public data we will describe the extraction precisely enough for somebody else to repeat it and disagree with our conclusions.
Where data comes from client accounts it would only ever appear aggregated, anonymised and with written permission, and the write-up would say exactly which parts cannot be shared and why.
Can our business take part in a study?
Possibly. Any study that depends on participation will say so plainly and will explain what taking part involves before anyone commits to anything.
Participation would never be a condition of working with us, and no client data appears in anything published here without explicit written permission for that specific use.
What would make you retract a study?
Discovering that the method was flawed, that the sample was not what we described, or that the conclusion does not follow from the data. Any one of those is enough on its own.
A retraction would stay visible rather than being deleted, with a note explaining what was wrong. Quietly removing a study other people have already cited is worse than publishing the error was.
Have a question worth a real study?
If there is something in your market that everyone asserts and nobody has measured, tell us. The studies worth running usually come from a practitioner who is tired of guessing, not from a content calendar.
Last updated · Reviewed by Zubair Afzal