Thank you @livepositionaltrader for sharing the detailed thought-process behind the NBFC-FD approach.
The previous post was underselling the amount of careful thought
that has gone into the investment approach into NBFC-FDs. 
Subsequently with all the details involved, being discussed in this topic-thread,
hopefully now anyone else reading this topic-thread in future,
will not go for any abstract approach to calculating risk based on their “gut-feel” of a brand. 
Also, here are a couple of potentially overlooked aspects in the “NBFC-FD” approach described so far.
( Sharing just in case these were not considered yet.
In case these aspects were already considered,
please do elaborate if you arrived at a different conclusion than the following. )
1. Comparing returns from NBFC FDs to LIQUIDCASE ignores their differences in liquidity.
What was the lock-in period of the NBFC-FDs offering the 7-8% annual rate of return?
For example, if we are considering NBFC-FDs that mature over 18-60months,
then the appropriate benchmarks to compare them against would be
T-BILLs, GSECs, SDLs, maturing over a similar time-horizon.
These assets offer higher returns than LIQUIDCASE,
- due to the lower liquidity
and higher expected value than NBFC FDs
- due to the further drop in probability of default (Sovereign vs. CRISIL-AAA).
2. The methodology behind the default-rate implied by a CRISIL-AAA rating is unclear.
Could you please share any references that led to the following assertions?
i would like to understand the methodology behind the same.
As an additional data-point, looking at CRISIL’s own report from 2022-23
(that popped-up in an online-search for CRISIL ratings and their associated default-rates)
which mentions that over a 10-year analysis period,
the cumulative default rates are significantly higher than 0.1% over a 2-3 year horizon
(the periods over which NBFC-FDs are offering the high 7-8% returns being discussed)
...and even higher default-rates during the recent relatively more volatile years.
[
Source ]
Rate of returns of an NBFC-FD over various tenures
[ Source ]
If we consider the range of default-rates from the above CRISIL report,
i.e. 0.07%, 0.16%, …, as high as 0.47% in recent times,
EVFD = (1 - 0.0007) × (1 + 0.0595) = 1.05876 (~5.88% risk-adjusted return)
...
...
EVFD = (1 - 0.0047) × (1 + 0.0595) = 1.05452 (~5.45% risk-adjusted return)
Note: The above math is overt nitpicking. The margin-of-error around these numbers is higher than what the above precision implies. The point being that, at extremes, even tiny numbers make a outsized difference.
PS: A previous draft of this post contained an error in calculations, despite my best efforts to triple-check it before posting. This has only further solidified my belief that the human brain cannot intuit accurately at extremes. 
Please do highlight if any other errors, either logical or mathematical, have still slipped through.
A 3rd overlooked aspect. (potentially off-topic if each investment is evaluated in isolation)
3. The entire capital invested in an NBFC-FD is at the designated risk
(…however minuscule the risk may be.)
If investing in such instruments to maximize returns, while…
- …lowering overall risk
- …ensuring predictable-income / future-assets
…then the barbell approach can be of relevance in this context/scenario.
Attempting to extract a few additional points of return
by investing in a “middle-of-the-road” somewhat riskier proposition
is usually worse than explicitly splitting one’s investments into 2, and,
- ,continue investing the fraction that one cannot afford to lose.
- at zero risk (as near zero in reality as possible).
- ,investing the fraction that one can afford to lose.
- at a significantly higher-risk (as high as one wishes/expects overall returns to be)
PPS: This barbell approach is also applicable outside of purely financial scenarios.
— your instinct to prioritize banks might actually be smarter (many ideas underneath) than chasing some stupid math. like beyond percentages and credit ratings, theres a critical layer we often ignore real-world usability.