Can I use estimated data in Bayesian updating?

Hi guys,

I am doing a Bayesian updating for housing construction defects. But the observed data is very limited and incomplete.

My model is beta binomial. beta (a1, a2) is my prior for defect rate for roof connections. The number of defects in n connections follow binomial (n, p), where p is the defect rate. My prior knowledge for this defect rate is 2%.

But the data is absent, and the only information I found in a thesis saying that: a house survey inspected about 100 houses, on average, each house has 100 connections and 3 are defective. But no details for the data.

My question is, Can i say the data I used for Bayesian updating is 100 roof connections inspected and 3 are defective, that leads to a posterior of beta (a1+3, a2+100-3)? Is it reasonable to use this hypothetical data based on limited information?