Create your own datapoints with Custom Data
Generated Custom Data lets you add your own question to a True Companies collection as a column. Beauhurst answers it for every company in the list and shows how it reached each answer. Use it for the checks you’d normally do by hand, like a scoring rubric, a fit test or a screening step. You end up with a consistent answer per company, saved where you’re already working.
Custom Data is in early access. It works on True Companies collections only. Each user can create up to five datapoints and gets complimentary credits to run them.
Before you start
You’ll need a True Companies collection. If your account isn’t on True Companies yet, speak to your support team.
Datapoints are generated from Beauhurst’s structured data on the True Company page, such as financials, grants, patents, fundraisings, people, shareholdings, news, industries and buzzwords. It doesn’t search the web or read the documents behind filings.
Create a datapoint
To begin, open a True Companies collection. Click 'Add' above the table, and choose 'Generate new datapoint'.
1. Write your question and pick an output type.
Describe what you want to know about each company, and give as much detail as you can. Then choose the type of answer you want back:
Output type | Use it for | Example |
Score | Ranking and prioritising a list | Fit against your ICP, 1–10 |
Classification | Sorting a list into named groups | Threshold met / approaching threshold / well below |
True/false | A straight include or exclude | Makes its own products rather than reselling or distributing others |
Number | A figure Beauhurst doesn't hold as a field | An estimate calculated from filed accounts |
Text | Open questions where the detail is what you need | Red flags in recent news coverage |
2. Review the refined prompt.
Beauhurst rewrites your question as a clearer, structured prompt. Read it carefully. It should be asking what you meant and pointing at the right data. Edit it if it isn't.
3. Preview on 10 companies.
Previews are free, so preview and refine as often as you need. You can't edit a datapoint once it's saved, so this is the time to get it right.
4. Choose how many companies to run it on, then save.
Select how many companies to run it on. You'll see the credit cost before you confirm. It runs on the companies at the top of the collection, in whatever order the table is sorted, so sort the list first if you want certain companies covered. When you save, choose whether to keep the datapoint private or share it with your team.
See how each answer was reached
Results appear as a new column in your collection. Click any value to open a panel with:
the reasoning behind the answer
the sources it used, so you can check the underlying data
who last generated it, and when
Use the arrows to move between companies in the list. To see the exact prompt behind a datapoint, open it from the column header or the reasoning panel.
If there isn't enough data to answer confidently, the datapoint returns an unknown instead of guessing.
Writing a good question
Name the data you want it to use. If your question refers to something Beauhurst holds, such as Innovate UK grants, patent filings, headcount, filed accounts, or Industries and Buzzwords, mention it by name.
Define your own terms. Words like "innovative", "good fit" or "genuine growth" mean something specific to you, but not to the model. Say what they mean in your work.
Be strict when you'll rank or segment on the answer. For a consistent result across hundreds of companies:
say what each score or category means
set the time period that counts
say what to do when the data is incomplete
The refine step in stage 3 helps you build this structure.
Keep it loose when the reasoning is what you're after. For open questions like adverse media screening, a looser prompt lets the model pull together more of what's relevant. Ask for the evidence and let it report what it found.
Example: innovation intensity score
Score the company on its innovation intensity based on R&D grants received, patent filings, academic collaborations, and whether it operates in science, engineering or deep-tech sectors. Give more weight to activity in the last three years. 1 = no innovation indicators; 4–5 = some recent small grants or initial patent filings; 8–9 = major R&D grants, significant patent filings and deep academic partnerships. Where data is incomplete, score on the evidence available and don't treat missing data as a negative.
Managing your datapoints
You can see and manage all your datapoints from the Custom Data page in Settings.
To add a saved datapoint to another True Companies collection, use custom columns, then run it there.
Shared datapoints can be viewed and run by anyone on your subscription. Only the owner can change its permissions or delete it.
Deleting a datapoint frees up one of your five slots.
Credits and limits during early access
Free to use with complimentary credits during early access.
Previews and failed runs don't use credits.
Credits come from whoever runs the generation, including when they run a datapoint shared by a colleague.
Credits aren't refunded when you delete a datapoint.
Each user can create up to five datapoints.
Generating data will become chargeable in Jan 2027, and limits will increase when it does.