Home/Before you hire
Before you hire anyone
The questions worth asking any data supplier
Including us. A supplier who cannot answer these is not necessarily dishonest, but you have no way to tell good data from plausible-looking data without the answers.
“What is your sampling frame, and who is excluded by it?” Almost every wrong conclusion drawn from a dataset traces back to that answer, and it is the question suppliers are least often asked.
What is the sampling frame, and who does it exclude?
Every frame excludes somebody. An online panel excludes people who are not online; a daytime household survey excludes people at work. That is not a fault as long as it is known and stated — the fault is a frame that quietly excludes the population your decision depends on. Ask for it in writing before any money is spent.
How was the sample size arrived at?
It should come from the decision: the effect you need to detect, and the confidence you need to act. A number picked to fit a budget produces a dataset that looks complete and cannot carry the claim you want to make. If a supplier cannot show the calculation, there probably was not one.
Who actually collects the data, and where?
Subcontracting is normal and not a problem; undisclosed subcontracting is. Ask which work is done by the supplier's own teams, which goes to a partner agency, who that partner is, and whether their cost appears as its own line or is folded into a margin.
What share of the work is back-checked, and how?
A share of completed work should be independently re-contacted and verified. Without back-checks a collection project is unauditable, and enumerator fabrication is a known and well-documented problem in field research. Ask for the percentage and the method.
What documentation ships with the dataset?
Protocol, timestamps, provenance, substitutions, response rates and limitations — prepared as the work proceeds, not reconstructed afterwards. None of this is visible in the delivered file, which is exactly why it has to be asked for in advance.
What is the lawful basis for collecting personal data?
If the data concerns people, somebody has to be able to name the basis and the jurisdiction. Ask which frameworks the supplier aligns to, whether they will sign your processor agreement rather than press their own template on you, and what the retention and deletion terms are.
What will you refuse to do?
A supplier with no published limits has not thought about where the liability lands, and it lands on you. Ours are published rather than discovered mid-project: no circumventing access controls, no scraping personal data for marketing, no fieldwork before ethics approval, and no conclusions the data does not support.
What happens if the answer is inconvenient?
Ask directly whether the supplier has ever told a client the data did not support the conclusion they wanted. The answer tells you what you are buying — a finding, or an agreement.
Do I own the result, and can I leave?
You should leave with the data, the configuration and the documentation, and ongoing arrangements should cancel with notice. Where work runs in your own cloud account under your own credentials, there is nothing to be locked into in the first place.
Ask us these
We would rather answer them before you commission than after. The scoping conversation is free and carries no obligation, and it is also where we say so if your question cannot be answered to the confidence your decision needs.
Ask for data
Requests and answers
No requests on the board right now. Yours would be the first.