What we shipped
- A data layer over our digital-PR operation that matches individual data points to the exact journalists who would care, drawing on eight years and 25,000 tagged journalist records in BuzzStream.
- Faster asset development: the moving infographics that are the backbone of our interactive data stories can now be built in minutes instead of days to the clients exact design specifications
- Currently live across a single template with more in the pipeline
- Automatic RAHP (Research, Angles, Headlines, Pitch) sheet drafting from datasets
- Press release outline builder based on the angles and datasets
Why we built it this way
We already had a productised PR framework in DISCO: discovery, ideation, sign-off, creation, outreach. Despite our very organised frameworks, there were still natural bottlenecks specifically around pooling data and matching it to the right journalist.
Unlike, thought leadership or reactive PR, data journalism has a much wider reach as different data points are relevant to different journalists. So whilst a thought leadership campaign would vertically stack a list of journalists around a single concept, data journalism stacks journalists horizontally across multiple concepts so you can infinitely expand outreach as much as the data will allow without sending the same headlines over and over again.
The setup and process are much harder and therefore the flop risk of campaigns not working is much higher. However, with our AI agent and PR machine, it reduces the flop risk by giving us much deeper lists of people to pitch with much more accurate data.
For example, the old way we would stop at 100 or so journalists as getting single lists of random journalists in a specific town, state or city was very labour intensive with a low payoff. However, we can now capture this long tail easily with an AI agent and a human in the loop to make sure the datapoint and headline are still newsworthy.
How it benefits clients
- Far more coverage, and better coverage. Instead of one thought-leadership piece sent to 50 loosely-relevant journalists, hundreds of tightly-relevant stories, each aimed at the one person it suits.
- Links from local and trade titles that rarely get fresh data, which now receive a genuinely relevant story with an interactive asset attached.
- More interesting assets for their social team to use and for their customer to interact with
- Un-scrapable content that won't end up consumed by AI as its presented as a tool and not flat content
What's next
- This is stage one. The PR Engine, a single data layer over journalists, outlets and opportunities, is in development to plan and run whole campaigns from one place.
- More templates and data visualisation and ways to present the stories
- A data ingestion engine to constantly source and scrape new data sources our clients can use for their campaigns.
Technical innovations
This unlocks the long tail of PR. Take a national dataset, say restaurant data across the United States for a pest-control client, generate hundreds of local data points, and match each to the journalist in that place, on that beat, who would want it. No human would sit and hand-match hundreds of rows to tiny publications, because the payoff per link is too small. A machine makes that economics work.
Now multiply this by creating as many assets as the datasets will allow on an always on publishing and pitching schedule using a single unified journalist database to make sure the same domains and emails only get contacted on a spaced out schedule with content that perfectly matches their beat.
We then have a PR 'machine'.
How it fits the agentic journey
This takes our existing DISCO framework and starts to mechanise some of the more tedious data collection and cleaning stages. It takes our AI manifesto of having a human hand at each turn and applies it to a traditionally very creative pursuit. By asking the AI to do more research and data analysis our PR consultants are freed up to do the thing that moves the needle - come up with new ideas and pitch them to journalists.