The lognormal distribution is typically used in situations where scope uncertainty is expected to be a significant factor in determining the actual cost.
A lognormal distribution extends indefinitely on the high cost side (ultimately at vanishingly low probability levels), but never goes below zero on the low cost side.
This distribution gives a higher probability of costs above the base estimate than below the base estimate. This is in accordance with the general experience that there are usually more things that can go wrong with a task than can go right, so the potential for exceeding costs is usually greater than the potential for coming in below cost.
There is no upper cost limit to this distribution, which again is in accordance with the general experience that every now and then a 'rogue' task can occur whose cost will be an order of magnitude or more greater than was budgeted.
Using the Lognormal Distribution
A lognormal distribution is defined in terms of a base or mid-range cost, represented by the peak of the curve, a contingency or high-range limit representing what you believe to be the probable upper limit of expenditure, and an overrun probability, i.e. a probability that the high-range limit will be exceeded.
N.B. Both the overrun probability and the contingency/high-range limit are 'built in' to the @Risk formula that Mandrel creates, and do not appear explicitly in the formula. If you change the overrun probability or the ratio of the high-range limit to the base cost, you must re-create the @Risk formula.