Work that doesn't fit the scrum model assigned to Scrum Team

For now I am just accepting these items will roll across sprints. 

Does the control of risk just roll across Sprints? Are there no unknowns that can be mitigated iteratively and incrementally? Is there nothing that can be learned about the product until the end?

One technique I've seen success with is something I'd refer to as "Milestone Splitting". Are there any logical breakpoints with the data load -> QA process where there might be a hand off, a double check, peer review, etc? 

If you identify the Milestones within the data load work then the team can demonstrate progress from Sprint to Sprint towards a larger goal that they know will span over multiple Sprints.  

Considering that you are working in life sciences domain I am assuming that you are dealing with tonne-load of sequencing data which might be result of NGS, Ion Torrent or Illumina. I doubt it is 454Pyro data.

My first suggestion would be to include an expert molecular biologist to the team to experiment your primary data loads. As most molecular biologists (honours or doctoral) have experience dealing and compartmentalising dataset, you will be able to find logical milestones as suggested by Tony. 

The next step would be check whether the data sets can be compartmentalized (not compressed) into functional entities viz. bp sequences, genes, chromosomes which you can measure and which of these is appropriate for your team to finish work in single sprint (if it is related and they are required to).

If this data is not novel you will probably be able to determine the appropriate function milestones by referring to GenBank or PDB (or of course your internal DB) whichever appropriate in your situation. 

If you are able to find these milestones, your team may be able to organize the work on Kanban or even in sprints. 

More than that, you will also be able to generate a metric that can provide you better estimates, value as well as help you prioritise.

I hope it helps.  Good Luck! 

Considering that you are working in life sciences domain I am assuming that you are dealing with tonne-load of sequencing data which might be result of NGS, Ion Torrent or Illumina. I doubt it is 454Pyro data.

My first suggestion would be to include an expert molecular biologist to the team to experiment your primary data loads. As most molecular biologists (honours or doctoral) have experience dealing and compartmentalising dataset, you will be able to find logical milestones as suggested by Tony. 

The next step would be check whether the data sets can be compartmentalized (not compressed) into functional entities viz. bp sequences, genes, chromosomes which you can measure and which of these is appropriate for your team to finish work in single sprint (if it is related and they are required to).

If this data is not novel you will probably be able to determine the appropriate function milestones by referring to GenBank or PDB (or of course your internal DB) whichever appropriate in your situation. 

If you are able to find these milestones, your team may be able to organize the work on Kanban or even in sprints. 

More than that, you will also be able to generate a metric that can provide you better estimates, value as well as help you prioritise.

I hope it helps.  Good Luck! 

@Emmy Chilton, I don't understand the technical aspects of the work you're doing, however, quite interested to learn more.

I work with a team that loads HUGE data files, the computer processing time alone can take 10 to 15 calendar days.  The data load itself is the deliverable so there is no way to break it down into smaller portions.   This is genetic info for a single species, it is already as small as I can get it.

How did you know the computer processing time is between 10 to 15 days? How did you get the ability to estimate? Once your team gets the data loaded, is that passed on to another team? is your work dependent on any other team and vice versa?

You've mentioned that the data load is the deliverable, therefore is there any human intervention and full-time effort required once the data upload has started all the way until it is actually loaded?

because Kanban is based upon resource agnostic work which is not the case for my team.  These data load requires expert level knowledge of Genetics to QA the info which not every team member has. 

Could you elaborate what you mean by "Kanban is resource agnostic" and how you came upon it? Also, is your team presently comprised of only experts in each knowledge area? Together, do they have all the skills to address the work?

Lastly, if you weren't using Scrum, Kanban or even trying to be Agile, how would your team/company have approached the exact same work/problem? Would they have been in a better position to predict delivery schedules? Why did the company choose to use Scrum or become Agile?

I'm not convinced that applying agile methods is appropriate for your situation. The agile methods are appropriate when developing or delivering complex products or services when there is a large amount of unknowns or uncertainty. It's useful in cases where you can't plan and predict how long tasks will take.

Are you able to look at the data files that you need to do and, with some kind of analysis, determine what steps are necessary and how long it will take to execute each step with some reasonable level of accuracy? If you are, I'd probably say that the agile methods aren't right and I'd consider a more plan based approach. You can add appropriate time to your plan based on the level of accuracy of your estimates and make reasonable up-front commitments to your stakeholders. After you get a schedule, you can look at your predicted dates and align that with the cadences of the team.

If you can't do the up-front analysis and there's uncertainty or unknowns, then there's probably a way to leverage agile methods (including Scrum). In this case, you should focus on two things. First, demonstrating progress on a regular basis (your Sprint cadence). Second, early indications when unknowns are going to impact your delivery because you have additional work or rework to do in order to meet the functionality or quality demands of stakeholders. I'm not sure if I'd call if Scrum if each Sprint doesn't result in something potentially usable by stakeholders, but you can leverage the cadence as an inspect and adapt cycle for risk reduction and quality improvement.