The small plan sponsors invariably want to know the plan cost attributable to each participant. For the Individual Aggregate and Unit Credit methods, this question is easily answered. However, answering this question for the EAN and the FIL methods has always been a challenge and problematic. I have used the following approach to allocate the Min/Max funding to each participant.

(a) Allocate Normal Cost in proportion to each participant’s EAN NC.

Plus

(b) Allocate S412/S404 net Amortization Charges in proportion to each participant’s UAL, where each participant’s UAL = Total UAL * Participant’s EAN AL / Total EAN AL.

However, at times the results produced are not palatable to the sponsors and are not consistent from year to year, even after allowing for changes in wages etc!

Your thoughts on and approach to this would be appreciated.

That technique is pretty common.
Also consider that you may have some amortization bases (for example, a plan amendment 3 years ago) which do not apply to all participants (for example, those hired after the amendment). Or, the credit balance is not associated with any new participants. Perhaps it would help to allocate such items to the participants "affected".

I am sorely tempted to ask: "What number would you like it to be?". But I usually don't. Especially in the case of a partnership or sole proprietorship where one must differentiate the deduction for purposes of income tax reporting.

Your method is as good as any other, because there is no such thing as a good method unless the plan uses a method where there are no bases and the costs are determined on an individual basis.

One could argue that the selection of a method that includes bases when the client will want or need a breakdown by individual is inherently inconsistent. But I wouldn't go quite that far. Close, though.

Whichever method you choose, since few people really understand defined benefit plans, I have found it necessary to clarify over and over to clients that this breakdown of "costs" does not work the same as an individual 401(k) account.

I actually had a client who tracked the annual cost breakdown used for deduction purposes and was irked that a participant's actual benefit was larger than this plus earnings.

You could also run a parallel IA method, and use the results to prorate your actual contribution.

I do all that and have started labelling it as Computed Funding Amount (not Cost) and even tell the clients not to look at and get hung up on the funding cost (which is function of cost method and assumptions used) but rather look at the PVAB of the benefit accruing during the year - which, ignoring the vesting, is the true cost of providing the benefit accruing during the year. But do clients buy that? NO!

For a fascinating example, see the first question at the end of chapter 1 of this book:
The Fundamentals of Pension Mathematics, by Barnet N. Berin

Can you summarize?

The example provides census data for a hypothetical plan, but essentially ignores all actuarial assumptions. Then calculate the funding contribution under several methods: Agg, EAN, UC, etc. All are equal, demonstrating that different patterns of funding are the result of applying actuarial assumptions, rather than applying a different method.

I did not believe it until I worked it out myself. Great exercise.

Even the frozen initial liability method?

Well, it must be, because it certainly doesn't make any sense to me!

Re-reading the book and exercise, I see that the exercise analyzed the Normal Cost and the Accrued Liability for 5 methods (UC, EA, EANFIL, AANFIL, and Agg.) No differences due to method!

The author concludes with "Note that all rational methods differ only becuase of the effect of actuarieal assumptions on the incidence of annual costs."

Without more detail I'm still a bit of a non-believer.