Reduce both to the same shape before comparing: one annual figure, over the same period, with the same things included, and the conditional parts separated from the certain ones. The most common distortions are comparing base to total, comparing target bonus to actual bonus, and counting a four-year equity grant as a first-year number.
Put both offers in the same units
The comparison fails when the two sides are not the same quantity. Base against total. Target bonus against what was actually paid. A grant against an annualised grant. Each of those is an unfair comparison and each is easy to make without noticing.
Build a table with one row per component and one column per offer, and leave a row blank rather than guessing at it. A blank row is a question to go and ask, and the questions it generates are usually the important ones.
Separate certain from conditional
Two subtotals: what you will be paid regardless, and what depends on something happening. Comparing the certain subtotals answers “which of these can I live on”. Comparing the totals answers “which of these might be worth more”. Those are different questions and conflating them is how people accept an offer they cannot actually afford.
Normalize the timing
A bonus paid in March to people employed in December is not available to someone starting in January. A cliff at twelve months means the first year of equity is contingent on staying past it. A review cycle two months after you start is worth more than one eleven months after.
Year one and steady state are different numbers, and for a role you might leave inside two years, year one is the one that matters.
The things that never make the table
Notice period, severance practice, whether the leave accrues and pays out, what happens to equity on redundancy. None of these is compensation until something goes wrong, at which point they are all of it.
Then stop
A carefully built comparison usually produces two totals within a few per cent of each other, and at that point the arithmetic has done its job and has nothing further to say. The decision is then about the work, the people and what the role does to your next move — which is the right basis for it, and which the numbers were never going to settle.
Common questions
What is the most common comparison error?
Comparing quantities that are not the same thing u2014 base against total, target bonus against actual bonus, or a multi-year equity grant against an annualised one.
Why separate certain from conditional?
Because they answer different questions. The certain subtotal tells you which offer you can live on; the total tells you which might be worth more.
Why does year one differ from steady state?
Bonus eligibility rules, vesting cliffs and review cycle timing all mean the first year can be materially different. For a role you might leave inside two years, year one is the number that matters.
What should I do about a component I do not know?
Leave the row blank rather than estimate it. A blank row is a question to ask, and those questions are usually the important ones.
What if the totals come out almost identical?
Then the arithmetic has done its job. The decision is about the work, the people and what the role does to your next move u2014 which the numbers were never going to settle.