A directed SBIR topic is a short document written by someone who has lived with a problem for years. Your job is not to be impressive; it is to demonstrate that you understood the problem the way they understand it and can solve it on their terms. Almost every avoidable loss on a directed topic comes from answering a question the author did not ask.
Decode what the author actually wants
Read the topic three times with different questions. First pass: what is the operational situation — who uses this, where, under what conditions, and what happens today when it fails? Second pass: what are the hard constraints, meaning the numbers, environments and interfaces stated as requirements rather than aspirations? Third pass: what does the Phase I deliverable list imply about maturity, since a topic asking for a demonstrated prototype in Phase I is a different topic from one asking for a feasibility study.
Then read the past awards under that topic number or its predecessors. They show you what the office has already funded, what it considers an adequate answer, and sometimes that your idea has been tried. If the agency ran a pre-release window, that conversation answers in ten minutes what the document leaves ambiguous.
Mirror the topic's language
Use the topic's own vocabulary for the same concepts. If it says "standoff detection," do not substitute your marketing phrase for the same thing — the reviewer is checking your text against the requirement list, often literally, and synonyms create doubt about whether you addressed it. Mirroring is not padding; it is removing ambiguity.
Then address every stated requirement explicitly, in the topic's order where the format allows. A short traceability table mapping each requirement to the section that answers it makes the reviewer's job easy, and easy correlates with scoring well. Where you cannot meet a requirement, say so and say what you will do about it. A stated limitation with a mitigation plan scores better than silence, which reads as either evasion or carelessness. Our piece on how review panels score proposals explains why an unaddressed line is treated as a failure rather than an omission.
The failure mode: pitching your product at the topic
The most common losing proposal takes an existing product, changes the cover page, and asserts that it satisfies the topic. Reviewers see this constantly. It shows up as pages of background on what you have built, thin coverage of the specific requirements, and a Phase I plan that is really a plan to demonstrate what already works.
If your product needs meaningful new engineering to meet the topic, say that clearly and scope Phase I around retiring the riskiest unknown. If it does not fit at all, skip the topic — the cost of a stretched bid is a cycle you cannot get back, and the base rates in SBIR success rates are unkind even to good fits. If the writing rather than the fit is your weak point, our take on hiring a grant writer is worth reading, and a pre-release call is covered in talking to a program manager.
Projects House writes the technical approach with you and builds what Phase I promises. Send us the topic through the contact form.