A supernova explodes, at a distance R from planet Earth.
We know the total energy emitted from the explosion, and we know that 90% of this energy is emitted as neutrinos. We also know the average energy of the neutrinos.
We can detect them using a CH₂ liquid scintillator, and the reaction ‾ν_e+p -> n+e⁺ (of known cross section).

We put some detectors (>2) on several places on Earth in order to identify the source via a triangulation. For simplicity, in what follows, we consider only 2 detectors at opposite sides on Earth (i.e. distanced by D_Earth=12000 km).
Goal: determine the azimuthal angle (wrt the axis formed by the two detectors) from a measurement of the relative delay in the arrival time.
What is the time resolution needed in order to measure a certain θ with an uncertainty δθ?
ΔR=D_Earth*sen(θ) and Δt=ΔR/c.
So, resolution δt=D_Earth*cos(θ)δθ. (In general, we cannot use the small angle approximation: the source can be anywhere.)

For simplicity, consider the neutrino flux as a constant during a burst time T (e.g. 5 seconds).
What is the detector mass needed in order to have the needed time resolution?
(I.e., in order to have enough events during δt to say that there is an unambiguous signal.)
A time bin (equal in width to the time resolution) is identifiable when the evenience of 0 counts is very improbable (let's say < 5% probability) given the average rate.
From Poisson statistics, P(0,μ)=e^-μ. So our condition is μ>-log(0.05)~3.
In turn, μ=Φ[ν*cm^-2s^-1]*δt*σ*M(detector)/m(CH₂ molecule).
Now it's just sufficient to solve this equation for M(detector), given μ>3.
To obtain the flux Φ from our initial conditions, take the energy flux as E_ν^tot/[T*4πR²] and divide by the average neutrino energy.