Data governance

How we handle your data

Most of our work can be done without ever seeing a record about an identifiable person. Where that is not possible, we would rather set out our position plainly before you ask than have you discover it halfway through an engagement. This page covers what we ask for, how it reaches us, who sees it, how long we keep it, and the point at which we will not proceed without an agreement in place.

Our position in short

We ask for the least data that will answer your question, which for a first-stage review is usually no participant-level data at all. Where identifiable or special category data is genuinely required, we will not accept it until a written agreement is in place that says what we may do with it. Material is held only for the engagement and deleted within ninety days of completion. ResearchBench is a trading name of Splash Solutions Ltd, registered in Scotland, company number SC851065.

Before anything is sent

We ask for less than you expect

A surprising amount of statistical work needs no personal data whatsoever.

Methodological review

A methods summary, a data dictionary and your aggregate outputs are enough to audit an analysis. No participant-level records are needed, and none should be sent.

Study design and sample size

Needs assumptions, not data. Nothing identifiable changes hands at any point.

Analysis of your own dataset

Here we do need the data, but almost never the identifiers. Direct identifiers should be stripped before it reaches us.

Minimising

Strip what we do not need

Remove direct identifiers

Names, addresses, NHS numbers, dates of birth, contact details and record numbers should be removed before sending. A study ID is enough for us to work with.

Keep your own key

If you need to link results back to individuals, hold the key yourself and never send it. We work with the pseudonymised file and you re-link at your end.

Reduce precision where it does not cost you

Age bands instead of dates of birth, and region instead of postcode, often answer the same question with far less risk attached.

Special category data

Health data, and where we stop

Information about health, and several other categories, carries additional protection under UK data protection law. We treat it accordingly.

An agreement comes first

We will not take participant-level health data without a written data-sharing or processing agreement executed beforehand. We are glad to work to yours, or to supply a draft for your legal team to amend.

Your approvals are yours

Ethical approval, participant consent and the lawful basis for processing rest with you as the organisation running the study. We will ask what they are, and we may decline if the answer is unclear.

We will say no

If an engagement would require data we do not think you are permitted to share, or that we do not need, we will say so and propose a narrower way to do the work.

In practice

Transfer, storage and deletion

Getting it to us

Email attachments are fine for anonymised and aggregate material. For anything participant-level we will agree a method with you first, and we are happy to use your own secure transfer or file-sharing platform rather than ours.

While we hold it

Material is stored on encrypted devices, kept to the people working on your engagement, and not copied to anywhere it does not need to be. It is never used to train anything or for any purpose other than your work.

Afterwards

Data is deleted within ninety days of the engagement completing. If you want it deleted sooner, or want written confirmation that it has been, ask and we will do it.

Roles

Who is responsible for what

Worth being precise about, because it determines who answers to whom.

You are the controller

In almost every engagement you decide what data exists and why, which makes your organisation the data controller. The study, the consent and the lawful basis remain yours.

We act on your instructions

We process what you send for the purpose you set, and for nothing else. That makes us a processor, and it is why the agreement setting out those instructions matters.

If something goes wrong

We would tell you without delay, in writing, with what we know and what we have done. You would then be in a position to meet your own reporting obligations.

Publication

Confidentiality and what we say afterwards

Your work stays yours

Findings, data and documents belong to you. We do not publish, reuse or refer to your work without written permission.

Case studies by agreement only

If we would like to describe an engagement we ask first, and we are content to describe it anonymously or not at all.

Reviews stay independent

Where we review work for publication we ask to be acknowledged as independent reviewers rather than named as authors, so the independence is visible to an editor.

Common questions

Answers before you ask

Do you need our raw data?

Usually less of it than you expect, and for a methodological review, none. A methods summary, a data dictionary and aggregate outputs are enough to audit an analysis. Where we do need participant-level data, direct identifiers should be stripped before it reaches us.

Will you sign a data sharing agreement?

Yes, and for participant-level health data we insist on one before anything is transferred. We will work to your agreement or provide a draft for your team to amend. We are also content to sign a confidentiality agreement separately.

How long do you keep our data?

For the engagement, and then deletion within ninety days of completion. We can delete sooner on request and confirm it in writing.

Who will see our data?

Only the people working on your engagement. It is not shared with anyone else, not used for any other purpose, and not used to train any system.

Are you the controller or the processor?

In almost every engagement you are the controller, because you decide what data exists and why. We act as a processor on your documented instructions. The agreement between us records those instructions.

What if we are not sure we are allowed to share something?

Tell us and we will work out what the question actually needs. Very often the answer is a reduced dataset, or aggregate outputs, that avoids the issue entirely. We would rather narrow the work than put you in a position you cannot defend.

Ask before you send anything

If you are unsure what we need, or what you are permitted to share, describe the project and we will tell you the least data that will answer it.