The survey is the story

Almost every digital PR campaign that earns coverage now runs on original data. The share sits at 95.9%, and the reason is not fashion. A reporter can get a quote from anybody with a title. They cannot get your number from anybody but you.

Australian newsrooms are smaller than they were five years ago and the people still in them are filing more per week than they used to. That changes what a pitch has to be. It is no longer an argument for why your company is interesting. It is a finished ingredient: a fact, a source, a chart, a spokesperson, all in one email.

Here is how to build one that survives a fact check.

What counts as a number

Four tests. A finding has to pass all of them before you spend a cent on fieldwork.

It has to be new. Restating an Australian Bureau of Statistics release is not research, it is homework, and the desk already has the release.

It has to be about behaviour or money. What people did, what they paid, what they stopped paying for. Attitudes are weak. A line saying 62% feel concerned about something is a sentence nobody reads twice. One in three delayed a purchase over $5,000 in the last six months gets picked up.

It has to cut locally. A national number gets you one story. A number that splits by state gets you the national story plus separate angles for the mastheads and radio in each capital.

It has to be defensible. If a reporter asks how many people you surveyed and in what week, you answer in one line without checking with anyone.

The build

1. 1.Write the headline before you write the questionnaire. Draft the sentence you want to see printed, then work backwards to the question that would produce it. If you cannot write that sentence in advance, you are commissioning a fishing trip.

1. 1.Size the sample for the claim you intend to make, not the sample you can afford. A nationally representative panel of 1,000 gives you a margin of error of roughly plus or minus 3.1 points at 95% confidence. That is fine for a national headline. It is not fine for Queensland business owners, because that subgroup might be 90 people. Any cut you plan to quote publicly needs about 300 respondents of its own, which means boosting quotas before fieldwork, not hoping afterwards.

1. 1.Weight to the Census on age, gender and state. Say so in the methodology note. Panel providers do this as standard, and stating it is what separates a research finding from a poll on your own mailing list.

1. 1.Write questions that cannot be accused of steering. One idea per question. No stems that assume the answer. Money questions in ranges, because people guess badly at exact figures and guess consistently at bands. Always include a none of these option, otherwise your percentage is something you manufactured rather than measured.

1. 1.Field it inside a week and record the exact dates. Someone will ask.

1. 1.Produce the methodology note as its own document: provider, sample size, field dates, weighting frame, margin of error. Attach it to every pitch. Half the reason research pitches die is that the journalist would have to chase you for this and does not have time.

The Australian cut

This is where most survey work leaves coverage on the table. One national release goes to the national desks and competes with everything else that morning. The same dataset, split by state, gives you eight distinct stories with a local number in each headline, and state rounds have far more space to fill than national ones.

Same fieldwork. Same cost. Eight more angles, plus talkback radio, which loves a state comparison and will book a spokesperson for it.

The sequence

Offer it to one masthead exclusively first, with a stated window of 48 hours. An exclusive gets a real journalist to actually open the file rather than skim the release. When the window closes, go wide with the full pack.

The pack is: a two-paragraph summary with the headline number in the first sentence, the data file as a spreadsheet with every question and every cut, two charts as image files sized for web, the methodology note, and a named spokesperson with a real title who can be on the phone within the hour. Include one line of genuine opinion from that spokesperson about why the number looks the way it does. The data gets you in. The opinion is what gets quoted.

Send Tuesday to Thursday, mid-morning. Not Monday, when the week's news list is being fought over. Not Friday.

What this does not do

It does not work if the finding is obvious. If the answer to your headline question is what any reasonable person would have guessed, you have paid for a press release nobody needs.

It does not work if you commissioned it to prove your product is necessary. Research designed to reach a conclusion reads as such, and the reporters who cover business have seen a thousand of them. The finding has to be allowed to be inconvenient for you. Some of the best coverage comes from a number that complicates your own pitch.

It does not work at small samples cut fine. Quoting a state result off 90 responses is the fastest way to get corrected in public.

And links are a byproduct. If you brief the work as a link-building exercise you will optimise for volume of coverage rather than quality of number, and you will get syndicated junk. Brief it as journalism you happen to be funding.

The number you own is the only asset in marketing that gets more valuable every time somebody else repeats it.