We usually think that bacteria are out there on their own. But they, like other single-celled species, can aggregate and act as a single organism under some circumstances (bacterial biofilms and slime molds, sponges and others are examples). Interestingly, the cells that make up the aggregate body are not necessarily those that shed to form another organism, such as sperm or eggs in mammals: even in a sponge, there can be separate 'body' and 'germ' cells. The body cells reproduce within the body but, like worker bees, are evolutionarily subordinate to the queens--the few reproductively active cells. Presumably, aggregation can at least have a collective advantage even if individual cells go their own way most of the time, and there is no evidence that I know of, other than chance, that determines which of the sponge's founding cell lines will end up as reproducing cells.
Cells in 'true' multicellular organisms, like humans, don't have any option of going it alone. We begin life as one cell, and develop into a differentiated organism with many types of specialized cells (organs in animals, roots and leaves in plants). Most of these don't reproduce, but the cells that do reproduce are genetically very closely related, so other cells aren't total evolutionary dead-ends. Even a super-organism like a bee or ant colony has only a subset of organisms that directly reproduce, creating representative descendants of the whole group.
Not only do we have specialized organs, but they are typically comprised of a great number of cells of different types. Bacterial species can specialize in many different ways, but the cells in multicellular organisms generally specialize only in one: they are intestinal lining producers, or muscle cells, or cells of the neocortex. That's a kind of cooperation within an organ analogous to the cooperation among organs that make the organism.
But there is a danger. Each cell division introduces mutations that will be carried by the cell and its daughter cells for the future life of the organism. The organ in which the mutation occurs is stuck for life with the mutant cells. The mutations are usually silent, or individually minor in their effects on the cell's behavior, but with millions or billions of cells in an organ, that has to work for a long time, at least some mutations may well have an effect, and some of those will be harmful.
A combination of such somatic mutations (SoMu) occurring over time may lead to a single cell lineage within the organ that no longer behaves properly, and in particular if it divides without the typical restraint for cell's tissue context in that organ, the changes can overwhelm the organ and that can threaten not only the whole organ but the whole individual. Cancer is the classic example in animals.
Given this, then why would organisms with mandatory multicellularity ever have evolved? Why not get together only when needed, as do the bacteria and slime molds of the world?
Safety in Numbers: Protection from mutational danger.
The cells in an organism do share a common genome, the one in the founding cell of the organism (fertilized egg, or seed). So an organism of varying specialized cells is a gang of likes, a differentiated, cooperating society of cellular kinship, which by aggregating can perhaps advance the cause of their group, their particular genotype, in a kind of Size Matters way: they can do things like exploit resources, just as bees and ants do, that an individual cell couldn't. Specialization, and size do make a difference. But the cost is that of the rogue members in the cellular society, whose SoMu number and sub-lineages increase with body size and age. When one organ fails, the whole organism fails.
One aspect of the protection of multicellularity is that SoMus will have various effects, from none to organ failure and death. Even if one cell lineage doesn't work efficiently, the organ itself is made of many other properly acting cells and even if an SoMu kills the cell, this may have no effect on the organ or the individual, with their countless normally behaving cells. A herd can withstand the bad behavior of a few of its members.
The risks that being a multicellular organism entail are offset by the average behavior of the aggregates of cells, and it usually takes time before any rogue sub-lineage would be life-threatening to the organ or its organism, as for example cancer is. Meanwhile, the organism can go about its business and take advantage of being a big, cooperative collective of organ functions, doing many things--travel, browse, hunt, mate and reproduce--in ways a single-celled organism can't do.
Mutations in parents, those arising in their genome and transmitted to their offspring, will either be selected against during development or will force the offspring to compete with its fellow organisms in the usual Darwinian way (see our series on the many other forms adaptation can take).
Since single-celled species, or those that spend most of their time as independents, are clearly doing very well and have done so for nearly the entire history of life (fossils of bacterial biofilms around 4 billion years old have been found). So multicellularity was never an overwhelming advantage, even if it opened different ways of life for some--and these relative exceptions are the most visible species.
The safety-in-numbers aspect of multicellular organisms seems to be a good way that being big can be successful even in the ever-present face of mutations, most of which are harmful. Safety in numbers may have allowed multicellular organisms to evolve in the first place.