Run a PR Test

A PR test shows what an announcement does to opinion. Paste a press release, a policy, a price change or a statement, choose an audience, and the run reports how opinion of the subject moves, where each group stands afterwards, and what people say about it over the following days. Use it before you publish, to compare wording or to test a response.

Set up the timeline

1. Open New Task, choose PR test, and select the audience.
2. Attach an announcement. Pick one you saved before, or choose New announcement. In the builder, paste the announcement exactly as people will see it, or use Import to bring in the text of a press release, memo or policy (PDF, Word, PowerPoint or text). Add up to four other wordings as versions B to E. Saying nothing is always in the comparison, so you never set it up yourself. Choose Finish to attach it.
3. Write the research goal in the composer, check the cost, then choose Run PR test.

There is nothing else to set, and no panel size to choose. Every audience is read the same way: 500 people give their view of the subject before and after the announcement, and 100 of them carry the discussion over five simulated days on a general social feed. What you might do on a later day is tested after the first result, against the same run (see Compare responses).

Before anything is charged, the run checks that the announcement is something people react to, such as an announcement, a policy, a statement or a post, and that every other version is a wording of the same announcement. A recipe, a questionnaire or a request to the assistant is stopped with the reason, and so is a version about something else. Nothing is charged for a run that stops here.

Announcements are kept in the Library under Announcement, so you can run the same one again on another audience or edit it later.

What you get

The card in the conversation keeps the key lines: which version works best, which groups end net negative, whether a wave of attention is likely, the overall change and the objection that spreads most. Each line carries a label, Strong, Good or Directional, read off the run itself: how many of the independent predictions agree on the best version, how far the overall change sits above its noise and whether the discussion moved the same way, whether a group's gap clears the noise of that group, in how many posts an argument was raised. Hover the label to see the figure behind it.

An audience with fewer than 500 people is read on the people it has. Below 300, the card shows no labels and no tracker points for that run, only the shares before and after, the ranking of versions and the discussion.

The detail opens in the Explorer on the right when the result arrives. Choose a line on the card to jump to the part that explains it, or Details to see everything. The Explorer has a section for each of the parts below.

Opinion before and after. Every person read gives their view of the subject before the announcement and after it. The card shows the share negative, neutral and positive at both points, how many people moved down or up, and the expected change on a tracker net score.

Which version works best. Add up to four other versions of the announcement. The card ranks every version by how it moves the answer against saying nothing, and marks any version that does worse than saying nothing. Compare versions of the same announcement for the same audience; this is the most reliable read the run gives.

How much attention it draws. For audiences in the United States and the United Kingdom, the card scores the announcement from 0 to 100 against past announcements and says whether a visible wave of public attention is likely. It shows the peak as a multiple of an ordinary day for the subject, with the range seen for comparable announcements. An ordinary day is the subject's usual level of public interest, measured by how many people look it up. The curve follows the usual shape of real attention: most of it on the day of the announcement and the day after, and back near normal within two weeks. Each later event you place on day 3 or after is marked by how likely it is to start a second wave.

Where each group stands. Switch between party, age, gender and race or ethnicity to see the share positive, neutral and negative in each group after the announcement. A group that ends with more people negative than positive is marked Net negative. The change at the top is for the whole audience.

What people object to. The arguments raised most often, with the share of posts that raised each one and a line someone wrote. Choose Write a version that answers this and the objection goes to the message box. Send it, and a new card shows a version that answers that objection, keeping every fact, price and date of your announcement, ranked with the others against saying nothing. This is free: only the ranking is redone. To see the groups, attention and discussion for the new version, choose Run a full timeline with it on that card, or Save as an announcement to keep it in the Library.

What people say. The discussion over the following days is kept in full: the arguments that appeared and when, the posts that spread, and the people who changed their mind and why. Use it to find the objection you have not answered yet.

How the method was checked. The version comparison picked the better wording in about 9 of 10 of 87 published randomized message experiments. The order of groups after an announcement matched published polls about 9 times in 10. The direction of the overall change was right about 7 times in 10 on 132 real events; read its size as a guide. The attention score told a visible wave from none about 8 times in 10 on 200 past announcements. The discussion raised about 6 in 10 of the arguments real people raised.

Compare responses

To test how you would respond, choose Test a response on the card. In the Explorer, write one step, such as "Day 4: the company apologises", or a whole plan of up to five steps, then choose Use this plan. The plan goes to the message box with its cost; send it to run it. It runs as a branch of the same run: the same people with the same first reactions, so only the response differs. The new card sets Without and With side by side for the change in opinion, and says which step is likely to start a second wave of attention. A response on day 2 is still part of the first wave, so steps start on day 3. A branch is charged like a run, one price per audience; a step after day 5 lengthens the simulation.

Choose Test another response on a branch card to try a different plan. Each plan is compared with the original run, not stacked on the previous plan. Differences between versions of your own announcement remain the most reliable read.

Cost

The charge is 800 credits per audience. Each audience is read on 500 people before and after the announcement, with 100 of them in the discussion over five days; the other wordings you add and the versions written to answer an objection are included. An audience with fewer than 500 people gives what it has at the same price, and a branch costs the same as a run. The composer shows the total before you run. Use reports to write up a run; the report opens with the change in opinion and where each group stands.