"We should dual-source that" is the easiest sentence in procurement to say and the hardest to fund, because the cost is immediate and specific while the benefit is hypothetical and vague.
It doesn't have to stay vague.
Expected annual loss
EAL = P(disruption/yr) × downtime_days × cost_per_day
cost_per_day is the part people underestimate. It isn't the value of the
parts — it's whatever stops. For an assembly line that means labour standing
idle, late-delivery penalties, and expedite premiums on everything downstream.
Worked, for a sole-sourced housing:
P = 0.12 (one meaningful interruption every ~8 years)
downtime = 18 days (their lead time plus your buffer, once buffer is gone)
cost/day = $4,200 (idle labour + expedite + one late-delivery penalty)
EAL = 0.12 × 18 × 4,200 = $9,072/year
Nine thousand a year is the premium you are currently paying, in expectation, to have one supplier. It just doesn't appear on any invoice.
What the second source costs
qualification = first_article + PPAP/inspection + tooling_duplicate
+ internal_hours × loaded_rate
annual_carry = price_premium × split_volume + audit_hours
Say $6,500 to qualify, and the second supplier is 4% dearer on the 30% of volume you'd move: 0.04 × 0.30 × $180,000 = $2,160/year.
Year 1 : 6,500 + 2,160 = $8,660
Year 2+: $2,160
Against an EAL of $9,072, it pays back inside the first year and returns nearly 7:1 thereafter. That is a fundable number, and it's the same shape as an insurance argument — because that's what it is.
Where the maths turns the other way
Run it on a low-value part with a fungible market and the answer inverts:
P = 0.12, downtime = 4 days (three shops can make it next week),
cost/day = $400
EAL = 0.12 × 4 × 400 = $192/year
$192 against $6,500 to qualify. Don't dual-source that. The instinct to protect everything equally is how resilience budgets get spent on parts nobody would have missed.
The two inputs that decide it are how replaceable the part is and what stops when it's late — not how much it costs.
The estimate everyone gets wrong
P(disruption). People reach for global events, but the common cases are
mundane: a supplier's key machine goes down, they take on a bigger customer and
deprioritise you, they have a bad quarter, they're acquired.
Your own history is a better prior than any published figure. Late deliveries,
missed quote deadlines and periods of non-response are all leading indicators
you've already observed — a supplier whose response rate fell from 90% to 50%
over a year is telling you something about P well before anything stops.
How this gets solved
Every term above needs history: lead times actually achieved, how often a supplier went quiet, whether a second shop ever quoted this part and at what price.
RFQTrace accumulates that as a by-product of running RFQs — response rates,
lead times, and every quote you didn't award. So P comes from your own record
rather than a guess, and the second-source premium is a number you already
have, because that supplier quoted the part last year and you kept it.