Sorry Tom,

Company confidential so can't send - I'll do my best to explain.

The data starts out hierarchically arranged which reflects the
structure of the source database (Borland Caliber - no I'd never heard
of it either).

We are talking mobile phone features here - A company will naturally
have a range of mobile phones all with a different set of features
(predominantly this is about software) that go to make up the unique
phone package. However, most features are used by more than one phone.
The database I am working with stores the data pertaining to these
features.

Each level 1 record is a unique Feature

Each level 2 record is a Requirement - a thing that is needed to make
a specific feature work, hence, in any given sort, it must stay under
the level 1 feature that it starts under and must only be sorted in
relation to other level 2 requirements (if there are any) pertaining
to that same feature

Each level 3 record is a Sub-component - generally one of many
predefined and tested routines that help to fulfill the requirement
that makes up the feature. Again, these level 3 records must stay
listed beneath their unique level 2 requirement and only be sorted
relative to other level 3 records listed under the same level 2
record. When a sort shifts the position of a level 2 requirement
relative to another level 2 requirement its level 3 subcomponents must
shift with it.

It should be noted that a sub-component may occur several times within
the datalist under different requirements and the same goes for
requirements (level 2) being used by more than one feature (level 1).
Therefore unique IDs cannot be relied on in any kind of solution here.

Typically a feature has two or three requirements each of which, in
turn, may have a dozen or so sub-components.

Status and Priority are just two of a number of column heading titles.
Each record, be it Level 1,2 or 3, will have a Status of anything from
Accepted to Rejected and a Priority ranging High, Medium, Low.

I hope this makes it clearer, I do appreciate you looking at this but
understand if it's too complicated to get into without seeing the
actual data. I'm reasonably confident that the solution I'm coding now
(described previously) will work. its just a case of how fast it'll
run and whether or not I can adapt it to a later task which will be a
4 level datalist (am I right in thinking Excel's limit is 17 decimal
places?).

Br, NickH