We increasingly engage in projects where we are asked to accommodate multiple collections, sites, and institutions into the planning, data modeling, and overall user experience. And we see a trend where grant funders actively encourage collaborations, so these kinds of digital projects may become common. It is important to think beyond the typical patterns of grouping sets of objects into institution-specific views, or presenting a mash-up as if it is just one big collection. As we think about collaborations involving online collections, we have identified human-centered user experience considerations and requirements to share with the community. It is common to see very carefully curated content such as catalogs for touring exhibitions. But open data and APIs create the opportunity to create less formally crafted applications. Designs must consider the degree of freedom that users can enjoy. What new pathways between objects become possible? Who creates – or discovers – them? How do users understand themes when they move among different museums’ works? Is vocabulary used in similar ways, or is there a "meta" descriptive language needed? Can different objects and media interact together, not just be experienced sequentially? And at what points does "where" matter – how do we give a sense of location for the physical works, without them feeling separated from each other? Our projects often focus on linked data as the underlying model, which can create new types of relationships between institutions' data. An important aspect of multi-museum collaborations is supporting staff as they prepare and review their information. Applications must provide insights into the scope of information, the degree of alignment between data sets from different institutions (dead-ends or conflicts that affect user navigation), and ways of identifying changes across the whole. These capabilities help all collaborators support not only resulting applications, but also each other.