You're right about being almost a CPAN module. But for the minimal case of encoding one string format to another, there is no reason for OOP.

If you provide a way to work with data sets in the abstract, an OOP approach starts to make sense.

• Encoder object keeps an array of series. Series are just arrays of numbers.
• Series can be managed using array-function like methods: push, pop, splice, etc.
• Provide a from_string method that takes decimal comma/pipe delimited decimal strings and makes objects.
• Provide a to_string method that generates comma/pipe delimited decimal strings from the object.
• Provide an encode method that generates an encoded string from the data series in the object.
• Provide a decode method that takes a maximum value and an encoded string and gets an approximation of the raw data.
• Provide a decode_scaled method that takesan encoded string and decodes it to values between 0 and 4095.

Note that you could just as easily expose a few functions and handle everything as native structures and go 100% procedural here, as well. OOP could be a nice win if you have a group of similar encodings that use scaling functions and serialization in similar ways. Then you'll actually be able to benefit from inheritance (at least if you designed your class to support the correct abstractions).

An even better choice would be to check CPAN and see if there are any multi-front-end charting libraries that could be extended to include this output format. What you do then will depend on the charting API's needs.

Oh, BTW, don't do bless( $self );, it silently breaks inheritance. Use 2 argument bless instead. The invoking class name is available as the first argument to your constructor. bless $self, $class;