How do I level tenders that are not like for like?

Levelling is the work of adjusting every return until they all price the same job, because until they do, comparing the totals compares nothing. Start with the qualifications and exclusions, not the price, because the bottom line is the least reliable figure in a return and hides whatever the covering letter carved out. List what each tenderer excluded or qualified, then either price it back in at a fair value so their adjusted number covers the full scope, or flag it as a risk you would carry if you appointed them. Check the arithmetic while you are in there, because a transposed figure or a rate applied to the wrong quantity can make a tender look cheaper or dearer than intended, and placing an order on their error just buys the argument later. Break each price back to the schedule's lines so the comparison is item by item, the only way to see a keen rate on one section masking a padded one on another. Where a tenderer has plainly priced a different scope, quantify the difference. What comes out is a table of adjusted true costs, and the cheapest adjusted number is very often not the cheapest headline one, which is the entire point of doing this before you commit rather than discovering it as variations once the works are under way.

Consider a main contractor levelling four tenders for a drylining package where the headline order runs from lowest to highest across a spread of a few per cent. Read casually, the lowest wins. Levelled properly, the picture inverts. The lowest excluded the fire-stopping the others included; added back at a fair rate, it jumps to third. The apparent second has an arithmetic error inflating its total, which corrected makes it the genuine lowest. The highest is the only one that priced the full scope with no qualifications, which on a job where certainty matters is worth something. The recommendation appoints the corrected second, having put the fire-stopping question to the first and a rate query to the fourth. Nobody was dishonest; the returns simply were not comparable until made so, and the number that looked cheapest left the most out.