Forum Discussion
Spread a value proportionally across several cells (with rounding)
I say 'ideally' as the policy is somewhat arbitrary, is subject to adjustment on a case-by-case basis, isn't being applied correctly at the moment, in 90% of cases the per-stage breakdown of days doesn't technically matter ... , and in the case of large projects the rounding errors are less significant. Currently we either run with the overcooked value, or manually adjust. Obviously automated is better, and as long as the overall project management effort adds up and *appears* to be spread roughly proportionally per stage, I'm good.
I'm not so sure, given what you wrote earlier in that same paragraph, that automation is obviously better. Especially if the number of projects is as finite as what you have shown in this series of messages. What could be automated (via some conditional formatting, perhaps) is the highlighting of discrepancies that need to be addressed, facilitating manual adjustment. Until such time as you've built AI into the process, being able to consider non-quantitative dimensions and make manual adjustments could be a good thing.
The current spreadsheet involves a fair few tabs, tables, named ranges, data validation, XLOOKUPS, pivot tables, one recorded macro with a button to update pivots (no other VBA), that sort of stuff... I'd guess it's Intermediate+ level excel, but there's nothing super advanced that a teammate couldn't understand if they put their mind to it.
The attached article is nearly two decades old, but contains some principles of spreadsheet design that you might find helpful. [I say this solely on the basis of your quick description, so you're surely free to disregard this and any other unsolicited advice.] Excel, as you know, is incredibly powerful. As such, it's not uncommon to see people designing things that mis-use (or even abuse) that power, making tasks more complicated than they need to be, designing tables and tabs that are based on the multiple ledger sheets or PERT/GANTT charts that were used in "the old days" before 1-2-3 and Excel. I don't know that this applies to you, but reading about good design is never an exercise in futility.
mathetes Well in the spirit of "mis-use (or even abuse) that power, making tasks more complicated than they need to be" I have attached a sample spreadsheet using both the LAMBDA I proposed and using 2 helper columns that basically re-do the functionality of that Lambda. So the 3 columns (2 helper and final) are:
25% down - this will take the 25% and round DOWN to nearest increment
formula: =IFERROR(MROUND(B2:B7/4-0.25, 0.5),0)
remainder - this will take the remainder after 25% to the nearest increment and applies a small offset based on the row number just so each should have a unique RANK
formula: =MOD(B2:B7/4,0.5)-ROW(B2:B7)/10000000
25% adj - this is the final value where if its remainder is in the top X rank it will round UP
formula: =G2#+0.5*(RANK(H2#,H2#,0)<=(B10-SUM(G2#))/0.5)
note in this workbook I added the 'distribute' function into the name functions so in column E you can see how that cell equation is simply =distribute(B2:B7,0.25,0.5)
m_tarler this is great, thanks! I can definitely use this (at least, I can use a version of the helper column version). I've spent my entire Sunday trying to take this apart and rebuild it so I can better understand it... with partial success...
I've dumbed it down a little (see Calc 2 in attached) by amending the formulae and breaking them out into as many separate columns as I could. This makes it easier for me to follow. If I can beg a little more help:
Does what I've done make sense?
Have I missed anything by swapping the MROUND function for FLOOR? (It seems a lot simpler. And I don't think I want the IFERROR wrap; this would cause an accidental minus input to default to a zero result, and I want an error message to be displayed instead.)This was my introduction to using # cell references with ranges and spills (thanks!!)... I've dumbed things down a little more in Calc 3 in attached) with dynamic arrays replaced with direct cell/range references for comparison. The #/range/spill versions are a lot more elegant than Calc 3, but are there any other strong reasons to avoid using Calc3? (The priorities are stability, simplicity, and performance, in that order. I suspect the dynamic arrays increase stability? NB I might be varying the no. of project stages in a future version but for now they'll be fixed at 6.)
Should I be converting this into a Table?
I'm planning to save the calculation in a protected worksheet to prevent accidents. I've had a test run at protecting the sheet and it seems fine, are you aware of any potential issues?
And now the annoying bit I didn't mention before! I'm trying to work around the following requirements, which mess up a very nice solution:- For historical and process reasons I should always have a value >0 (ideally a minimum of 0.5) in the final column for the first row/project stage.
- IDEALLY (!) I should always have a value >0 (ideally a minimum of 0.5) in the final column for ANY row/project stage where engineering days > 0.
The additional purple-highlighted workaround columns fulfil these requirements, but the result gets progressively less accurate with small no.s of engineering days in each row, as the rounding errors creep back in. NB:
- If total engineering days <=5, a different policy kicks in and I can apply an =IFS to apply that policy instead of the ~25% policy, making the problem go away.
- But for example with total engineering days =5.5, with engineering days in every row, I'm back to square 1 with the rounding issues (see example saved in Calc 2).
So... I'm considering a couple of workaround options:
- Manually force the Row 1 result to be 0.5 days (it'll never be higher than this anyway), and allow some other row results to round down to zero.
- Change my per-row rounding to 0.25 increments. As long as Row 1 >0, and as long as the overall total rounds up to a multiple of 0.5, I should get away with this.
I suspect other options/variants are available and may be easier. What would you recommend, and how would you adjust the xls to implement it?
Again, I really appreciate your help (and time), thanks.- I will try to answer best I can.
I think FLOOR is an improvement. I forgot that it will do multiples (I always think of it for integers)
I think using the # reference is better because it will auto expand/contract if the spill range changes. also note that excel will automatically replace a selected range that matches a spill range with the # notation. But in the end I see it has no other improvement and a direct reference is fine (until the size changes)
as for making it a table I don't think you will be able to with spill equations. you will have to change them to not spill
Protecting a sheet is good idea when other will be using it and you don't want them to accidentally break it. But not that the protection, no matter how clever you tweak it, should never be used to hide or protect any truely confidential or private information (I don't think that is an issue here).
an alternative way to make a min value is using MAX( [min val] , [value] )
yes as soon as you start requiring round up you will have problems (that is where this post started). and I've never been a fan of "never" but you could make the first one a MIN of 0.5 and then let the rest fall as they may. You might also be able to encourage the 0s to inc up to 0.5 over others by adding a conditional onto the remainder column that if it is a 0 then add 0.5 to the remainder and then those will always be prioritized over the others but if all are 0 then it will prioritize the 0s that are larger and then finally in order from top to bottom. And finally going to 0.25 increments might help but the rules will still apply and probably still be needed - basically yes the round up will be less but the undershoot is less also and you will still need on average 1/2 over and 1/2 under.
