Hi,
I have recently been trying to adopt to a more smoother workflow when I design new things in SolidWorks.
Therefore I would like to hear how other people work, and what they would consider as pros and cons regarding top-down and bottom-up design.
I tend to do bottom-up design when I create new parts, and then later on in assembly mode I turn to top-down.
All aspects of 3D modelling require forethought and discipline. Whether you design top-down or bottom-up, you need to have a vision and a plan.
Roland, you have asked a question that can only adequately be answered with a tome as large as War and Peace.
I personally ascribe to the bottom up paradigm, unless you can complete you design using Weldments.
Knowing exactly what can/will change/be revised later is extremely valuable information.
If you can, a skeleton sketch paradigm works great. Ask John Stoltzfus.
Mating planes/axis and never deleting anything from your assembly also helps.
Good Luck and let us know how/what works for you.
I didn't ask. I answered (sort of).
About the only time I use top down is for virtual parts, such as cable, that won't be detailed in the Drawing.
I thought I was the only one who uses virtual parts. I never see anyone mentioning it here.
I'm glad I'm not alone....
In my case , the components that won't be used in other projects/assemblies are virtual and I use top/down work flow.
I use bottom-up for the parts that are saved external to assembly.
Because I hate to use a part in an assembly which its size depends on another assembly.
I do not use top-down very much at all. In certain circumstances it can be very beneficial. It can get problematic if you create a part from a top-down assy and use that same part in another assembly. If your references are not broken and you change something on the part in one assy it will update the other assy which you may not want.
Much like Glenn Schroeder, my primary use of Top Down is with hoses or cables. I will occasionally use Top Down if I have known locations for items such as the front and rear end sheets of a mixer, or the scale mounts need to be is certain locations. However as Jeff Holliday and others have pointed out I prefer to go back and break the external references once done so I can use the parts in other assemblies.
On the flip side, if the parts will only be used in a certain assembly but different sizes or configurations of the assembly will be used in the future it can be VERY helpful to do a top down and keep the external references. A couple of quick changes to the design table and you have a whole new assembly to give the shop floor. Very impressive and cost effective.
The bottom line is both methods have their uses so it is beneficial to know both. You just need to know how the normal opperating procedure of your company and how you will be using the assembly down the road.
Top down or bottom up, That I feel isn't our choice when we design something, there so much that drives the complete assemblies, the workflow, does your company make everything or do you sub out a lot of components. First I would like to clarify my definition of "Design" - Something New, not Something Copied...
I feel the workflow you choose has it's own advantages or disadvantages - the problem with Top down, it takes more up front work, designing in context is a challenge, but I am more comfortable going that route because of the need for parametric assemblies, I really don't have much choice here.
The biggest problem with bottom up design is that it is very easy to create circular rebuilds which are a killer, plus if you have a complex assembly it is easy to over look details, because you're designing the component off to the side and bringing it in, so what happens when you need a round widget instead of square, try that, go and make that change.
Designing is all about ease of change, not ease of design, because design something "New" is bound to look entirely different from beginning to the end, paper napkin to final product. As near as I can tell based on previous experience and asking around there is about a 4:1 ratio between preliminary models and final design, 4 designs will be thrown away and you keep one.
I know it's a personal journey, everybody has their own way of doing things, but I like to challenge anybody designing, when their design is complete go to your feature tree of the final assembly or one of the sub-assemblies and delete a part and tell me what happens...... You should only get mate errors of missing faces, edges or vertex's - if you do it incorrectly the entire feature tree will turn into a forest fire..
Also, if you spend a lot of time up front in the "Design Intent" there will be less issues down stream...
In my work (consumer electronics), the approach is sort of bottom-up then top-down. Typically there is an overall design provided by industrial designers. We arrange key components and resize and redesign the ID as required. Then we shift to top-down for the enclosure.
It depends. Some parts are standard size and should be designed bottom-up to utilize time. Other parts require you to have a thorough understanding of what is already in an assembly to design it - top-down is best in this case. i.e. ductwork/engine block etc. If you are worried about references then make sure to select the option 
Overall I really like top-down designing. You just have to keep in mind the part will most likely not be orientated right on the origin, or on the desired plane you want. You can modify all of this, of course, after the fact. Then, delete the in-place-mate and re-mate (is that a word?) the component as needed.
cheers,
Nick Reak
I typically use top-down for new design work, but it includes a majority of bottom-up design work. It is too easy to go crazy with top-down relationships as you are designing something new. This can cause a lot of problems later on when you need to make changes, and you can end up with circular references. So, choose wisely and make a few key top-down relationships then build bottom-up details around them. Often, I build each component in it's own relative coordinate system, including it's main locating features, then add it to the assembly and add in-context relations if needed for the remaining major key features. That's a preference of mine, but I found it helps tremendously later when creating drawings.
Like some others have said, try to build on a framework/skeleton in your assembly (sketches, planes, etc), but also don't add small components like hardware and fasteners until the end. In other words, build your assembly by mating major components together or to sketches or planes in your assembly so they move properly without the small stuff - mate linkages with hole-to-hole relationships. As you build, ask yourself, "how will this assembly behave if I suppressed, removed or replaced this fastener/hardware component? The assembly should still move and behave the same afterwards.
Also, I use virtual components for new designs. It speeds up the creative process because I don't have to think about what to name the components and where to name them - I can do all that later, after my new design has solidified a bit. After my new design has stabilized, I then rename the virtual components, then save them to external files and replace all the in-context references with internal references.
I follow the same as others here.
If a complex design where all surfacing is involved, then I start with 3D sketching, top down design.
Otherwise, everything bottom up.
I use top down when it makes sense, it only takes one assembly to blow up on you before you become gun shy on using it.
Just use it as a tool when required and you will be fine.
At my last job the previous cad guy made all of his models top down and they all were garbage. He said he did that so he could scale one or two dimensions that affected the entire assembly however once you did that they blew up.
Another thing I have learned to hate is external references and equations, I have come across models that were useless to me because of that.
Only the person that created them knew how to successfully use those models. I think he did that for job security.
Several years ago I had a side job for a previous employer where I was given a model and asked to make several changes. I ran in circles for a while because I would make a change and everything else would change. Finally realized he had made it top down and removed the external references to get the job finished. Project took a lot of extra time to get done. If I had known what was going on up front I could have finished the project in less than half of the time.
The main reason Top down approach doesn't work for a lot of people is the simple fact that there isn't a common denominator controlling it such as a Master Model or a Master Sketch. If you build top down just like you built a tower of blocks when you were a kid, as soon as you take out or change any of the blocks mid way or wherever in the tower, the crash will be the same.
The key in good top down design is using a Master Model/Skeleton Sketch approach and when you add a new part link it to the Master, that way if and when you pull out any of the blocks in the tower the rest stay put. Any time you insert a new component the key is to Isolate the new part and the Master Part/Sketch, that way each component is independent but tied to a common structure..
So my point is either way you create your models, if your references are screwed up then your model is static and will blow up and very difficult to change, I hear guys creating models and then deleting the references, I sure hate to edit those models as well.
In my case, if there had been a master model or sketch I would have been able to recognize it and would have been able work with it. Instead the driving sketches, yes plural, were within a couple of the sub assemblies. I was 3 weeks in when I found them.
I think I would enjoy my beer better this way actually. Good perspective, haha.
I've liked the master model approach. It gives a pretty good blend of bottom up and top down. Mark Biasotti gave a really good presentation on it at SWW2017 "Master Model Best Practices".
My recommendation is to only use it when necessary. It's a nice tool, but the in-context relationships can be tough to manage when you have a lot of them. And they can slow the assembly down.
Sometimes I use top down during the design process to make sure all is lined up when I make frequent changes until the design is settled down. Then I delete just about ALL of them in the end.
Don't get me wrong, it can be incredibly useful, but also dangerous.