>> How much data are talking about?
>1 or 2 thousands max with many relations on 4 or 5 other tables of hundreds
>of records.

I meant how much data as in how many possible characters are you sifting through within the memo field. From your answer below, it looks like it can be a significant amount and it's variable.

>
>> If it's in the same position, can you add some post-processing to your save
>> routine to strip this out and pop it into a text field that can be easily
>> indexed and queried?
>
>It is the invoice detail and different product numbers may appear...

Index values have a maximum length of 240 characters (if a CDX is used, 100 if not) so it won't really help you to index on a subset of the memo field if the product numbers aren't contained within an area spanning that many characters. You are probably going to have to implement a full-text querying engine (e.g. PHDbase) or use $ or AT() in your queries (making them very slow).

This may seem academic at this point but do the product numbers have to be stored in a memo field? If they were in a child table, the product numbers could be simple text fields that could be queried. Why the memo field approach?