I worked for a long time at a leading high-quality antibody company, so I'll try and share some of my experiences with you. The process of making a highly specific antibody (I'll focus on monoclonals) has three important parts - antigen design and immunization, cloning and subcloning, and screening/validation. Each part is crucial on its own, and the better one is performed, the higher your chances of success downstream become.
In order for an antibody to work, it needs to bind specifically to its target. I won't get into all the specifics here, as there is a good open literature on antigen design, but basically you need something that will promote a strong immune response in the host, give you a variety of clones to choose from, and be specific - it shouldn't bind other targets besides your protein of interest. There are lots of factors to consider, including antigenicity, solubility, ease of production (for immunizing and screening), steric considerations (does another protein or nucleic acid obscure the target site in vivo?), and others. Antigens can be roughly divided into two types - peptides (typically less than 20 or 30 amino acids) and proteins or protein fragments. You will most likely want to try several antigens to maximize your probability of success. After the antigen is synthesized and purified, you need to come up with an immunization schedule for priming and boosting the animal(s), then determine when and how to test them (screening), and at what time and how to harvest it for the monoclonal process.
The cloning process then comes next. Clones can be generated by a variety of methods, some widely available, such as the various hybridoma methods, and some proprietary to individual companies. My former company employed a mix of both, depending on the species of animal, using some really cool in-house tech that was constantly being improved and expanded upon. These processes are very dependent on the quality of materials being used and the expertise of the cloners. The input B cells (which make the antibodies) need to have been harvested, treated, and stored according to a set of precise steps to allow for the highest number of viable clones. Once the immortalized cells have been obtained (after fusion with the myeloma partner cells in the hybridoma process, for example) they are diluted to a pre-determined cell count and plated in 96-well plates and allowed to recover and grow for a time, during which they are secreting antibodies into the medium. This media is then tested quickly (before the cells overgrow the well), and positive wells are diluted out to (hopefully) single-cell suspensions and subcloned, or frozen down for later. The first testing is frequently by ELISA using the immunizing antigen, although depending on your application of interest it may be by high-throughput screening via flow cytometry, immunohistochemistry, or immunofluorescence. After the initial testing, subsequent steps of subcloning and re-testing can occur at a more measured pace, as the cells can be frozen indefinitely after generating sufficient quantities of antibodies in the supernatant.
This is where the quality of your screening program comes in. Your assays need to be well-designed, robust, well-documented and performed, and have appropriate positive and negative controls so that you can truly tell whether a certain clone is of interest, or performs better than an existing product. This means controlled conditions, standardized reagents and protocols, and a high degree of repeatability from assay to assay. Also, the appropriateness of your controls cannot be emphasized enough. Do you have a knockout or non-expressing sample for comparison, or a way of looking at baseline vs. induced or inhibited activity? For ChIP, unless you have a good control antibody and primer pair for your gene of interest, how will you know if the assay works? You may get a great result doing a straight immunoprecipitation-Western blot, but still not pull down DNA-bound protein. If your ChIP method doesn't produce clean DNA, you may never get binding. And if you don't have appropriate data analysis protocols in place with multiple replicate wells, you may waste time chasing noise instead of specific signal. Additionally, along with testing all your clones, you also need to titrate them to determine the optimal working concentration. I've seen a lot of antibodies that look like crap when you first test the supe, but diluted 100 or 1000X they are clean, specific, and still strong. Be patient, testing is a lot of work and requires a ton of repetition to confirm your results, but it'll be well worth it in the end.
Finally, after all this is done, you hopefully have a great clone that does what you want it to do, and better than anything else out there. You now need to decide what you're going to do with it, because if you're an academic lab and publish with it, the world will come knocking at your door. Don't slack off at the end and decide 100 ml of supernatant will be enough to last forever - save the clones and put them in a cell bank, or try to get a commercialization deal with a manufacturing company so you don't have to do all the work!
I hope this helps, please let me know if you have any additional questions or concerns.