See Tom Jones, "Inferential Genealogy" course handout, 2010, Family Search. The goal might
be to identify who, from several known candidates (and possibly some unknown ones), is the parent of X. Or it might be to find some information out about a specific person. The goal should be specific in identifying people - it is not helpful to simply say "Identify the mother of X", you need to say something like "Identify the mother of X who married Y in place P in year Y". Even if you only have one person of name X now, you'll probably get another in the future, at which point your goal will become confusing if not specifically focussed.
I use the term "goal" rather than "objective" because I associate the term "objective" with "specific, measurable, agreed, realistic, timed" (SMART). At this stage, the goal might not be realistic. And I will use "Objective" later on for a lower level concept.
This is "short-hand" for any attribute of any person, family, group, place, ship, artefact, etc. entity. Or any event or relationship involving the same entities. The value includes the date(s) relating to any event or attribute, place, value of an attribute, any associated note, any role, any identities of related entities. This process does not define, nor depend on, whether or not a date and a place of an event (say) make up one value or two. That decision is a feature of the data model concerned and how values have citations recorded against them. The research justifications (i.e. proofs and conclusions) must appear at the same level as citations. In other words, if a citation can only appear against the entire block of date, place, value, note, etc (as per current GEDCOM), then the research justifications must also appear against the entire block. If
in BetterGEDCOM, separate citations can appear against the date and place (say), then it must also be possible to record separate research justifications against the date and place. This is for consistency.
Each work portion contributes a step towards the overall goal. Each work portion will have its own specific objective, which is lower-level / more detailed / more specific than the overall focussed goal. Note that I propose that the portion should have one
single objective that is specific, measurable (generally - how do you know you've completed it or not?), and realistic. Or, as realistic as it is possible to be without further research. The necessity for one objective is imposed just to keep things simple - "Who were the two parents of X who married W in year Y at place P?" gives rise to multiple entries in the database but can be easily expressed as one objective. "Who were the two parents of X who married W in year Y at place P and when did they marry?" is a single sentence but combines two objectives as the work for the marriage date can probably not start until the two parties have been identified so is (at least) two objectives.
A work portion should have a clear starting point - in other words, if you don't know whether you can start a work portion, then something is not properly defined - that something may, of course, be a previous portion. It should have a clear end point - otherwise how do you know if you've finished it? Whether those start and end points need to be explicitly written down is debatable. If they aren't, then they must be "obvious" from the text.
Subject to the requirement to have an objective, it is entirely up to the researcher what goes into a work portion as it is a conveniently - and perhaps arbitrarily - sized piece of work. There is no necessity that (for example) a work-portion finishes with an update of the database or a proof of anything. For instance, the first work portion may well be self-education such as "Educate self about where to find records of the Royal Navy in Nelson's era". Others may prefer never to have such work portions in their database. Also, some researchers may split work into several work portions where other would see only one. Provided the objective can be written out clearly in both cases, this does not matter.