Join errors · assumptions ·. The processor role
Professional indemnity insurance (BAV) for a data analyst
Your analysis disappears into someone else's decision. If there is a mistake in a join or in an assumption, the loss has already been suffered before anyone checks the figure.
- Several insurers compared objectively
- 9.5 customer rating for a new policy
- AFM licence 12016589
- Personal 072 - 509 24 56, weekdays 9–17
Deze pagina in het Nederlands: Beroepsaansprakelijkheidsverzekering voor een data-analist.
The calculator and the quote form below are in Dutch. Prefer to do this in English? Email info@finassverzekert. nl or call 072 - 509 24 56 and we will take it from there.
Work out for yourself what it would cost.
Enter your details; you will receive a proposal within one working day.
- We compare the offerings of several insurers
- An adviser checks whether the cover suits your activities
- We arrange the switch, including cancellation
A premium indication, not personal advice. Prefer to talk it through? Call 072 - 509 24 56.
In brief
Professional indemnity insurance (BAV) covers pure financial loss: loss that does not arise from injury or damage. In analytical work that is the whole risk. A join on the wrong key that leads to double counting, a filter that quietly removes part of the rows, a model trained on an unrepresentative period, a dashboard on which a client steers his purchasing, pricing or staff planning. The defect often only becomes visible after a quarter, and by then the decisions have been taken.
The standard is the duty of care in Article 7:401 of the Dutch Civil Code: to act as a reasonably competent and reasonably acting contractor. That includes naming the limitations of the source data, recording assumptions and reference dates, and keeping your scripts and versions. Reproducibility is your strongest defence when a claim arrives: without version control you cannot show which result you delivered at the time and which data it was based on.
If you work with personal data, you are almost always a processor within the meaning of Article 28 of the GDPR, and a processing agreement is compulsory. A fine from the Autoriteit Persoonsgegevens (the Dutch data protection authority) is not insurable. The cost of a breach or a ransomware attack in your own environment belongs on a cyber insurance, not on this policy.
This page deals with one situation. The full overview is on Compare professional indemnity insurance (BAV).
What to look out for
Four situations in which an analyst runs into trouble sooner than through an arithmetical error.
Double counting and silent filters
The classic loss does not arise from a wrong formula but from a join error: records counted twice, rows lost through a join, a time zone or currency that has not been harmonised. The cost of running your own analysis again is not covered. The consequences at your client are, in principle.
Assumptions are read as promises
A statement about the accuracy of a forecasting model quickly becomes a promise in a steering committee. If a claim rests solely on such a promise about the outcome, it falls outside the cover: what is insured is the professional error, not the promise. Put margins of uncertainty, the period used and the period of validity in the document itself, not in an email.
Data you were not allowed to use
Scraped sources, datasets with a licence that prohibits onward supply, or data the client was only allowed to process for another purpose. Claims about infringement of intellectual property and about breach of licence conditions are excluded on many policies or call for a separate section.
Who interprets, and who decides
If you deliver a dataset that the client interprets himself, the line lies elsewhere than when you write the conclusion. If the client ignores your reservation or adjusts the model himself, contributory fault under Article 6:101 of the Dutch Civil Codecomes into play. Record the handover and the final version in writing.
What does your premium depend on?
- Annual turnover: the usual basis for advisory and analytical work
- The sector your clients are in: healthcare, insurance and banking weigh more heavily than retail
- Working with personal data: the processor role and sensitive categories increase the risk
- Model building or reporting: predictive models count differently from periodic reports
- Area of cover: work for clients outside Europe is accepted separately
- Sum insured and excess: per claim and capped per insurance year
Insurers weigh these details differently. That is where your saving is.
What is covered
| Situation | AVB | BAV |
|---|---|---|
| Your forecasting model sets a client's purchasing a quarter too low and he loses turnover | No | Yes |
| You knock over a cup onto your client's monitor during a working session on site | Yes | No |
| An export containing personal data goes to the wrong recipient and your client has to inform the data subjects | No | Provided that |
| Your script runs on in the production database and a day has to be restored from back-up | No | Yes |
| Your client wants his fee back because the dashboard turned out to be unusable | No | No |
| Your own laptop with all your analyses is stolen from the train | No | No |
The distinction lies in the type of loss: damaged equipment belongs on the AVB, wrong results and the decisions based on them belong on the BAV.
Frequently asked questions
This is what people ask us most.
I have lost a client's data. Which policy is that?
Data is not property in law, so losing it is usually not damage to property but financial loss. The cost of reconstruction at your client therefore falls on this insurance sooner than on public and employers' liability insurance (AVB). If the cause is a break-in in your own systems, cyber insurance comes in as well, for investigation, recovery and the notification duty.
Is a fine from the Autoriteit Persoonsgegevens insured?
No. Administrative fines are a penalty and not compensation, and are therefore excluded on virtually every liability policy. A fine imposed on your client and passed on to you generally falls outside cover too. What is insurable is the loss that data subjects or your client themselves suffer through your mistake.
My client has changed my dashboard himself. Am I still liable?
Only for your own failure. If the client changes the logic or adds sources, the cause shifts. So keep the version delivered, the accompanying documentation and the moment of handover. Without that record it can no longer be shown afterwards which part of the model came from you.
A mistake from 2023 is only coming to light now. Is that covered?
Under a claims-made policy what counts is the moment the claim comes in and is notified. Your current policy only covers old mistakes if there is retroactive cover and the retroactive date goes back far enough. Circumstances you already knew about when you took out the policy are excluded. So when you switch insurers, always ask about the retroactive date.
Read more
Within Beroepsaansprakelijkheidsverzekering
- Beroepsaansprakelijkheidsverzekeringthe main page
- Beroepsaansprakelijkheidsverzekering ai adviesbureau
- Beroepsaansprakelijkheidsverzekering cybersecurity consultant zzp
- Beroepsaansprakelijkheidsverzekering cybersecurity bedrijf
- Beroepsaansprakelijkheidsverzekering data analyse bureau
- Beroepsaansprakelijkheidsverzekering ai consultant zzp
- Beroepsaansprakelijkheidsverzekering hr adviesbureau
Similar pages
- Professional indemnity insurance (BAV) for a data analysis agency
- Professional indemnity insurance (BAV) for a cyber security consultant
- Professional indemnity insurance (BAV) for an AI consultancy
- Professional indemnity insurance (BAV) for a sustainability consultancy
- Professional indemnity insurance (BAV) for an HR consultancy
- Professional indemnity insurance (BAV) for a self-employed AI consultant
- Professional indemnity insurance (BAV) for a cyber security company
- Professional indemnity insurance (BAV) for an IT consultant