Run the entire sales loop once against a real prospect, end to end: pick a plausible buyer for your practice product, research them in a timed 20-minute sprint, send a four-sentence cold email to a verified address, send one bump three business days later, and read the result. You submit the actual artifacts - the brief, both sent messages, the log, and a retro - not descriptions of them. The goal is one complete, honest rep of the loop, whatever the outcome; a well-run attempt that ends in silence scores higher than a sloppy one that got lucky.
Pick a prospect who is plausible but not precious - someone where a clumsy first attempt costs you nothing but the practice. Run the research sprint with a real timer and stop at 20 minutes; the brief is graded on specificity, not length. Hold your second-best research fact out of email one so the bump has something genuinely new to say. If you hear nothing after the bump, that is a normal, submittable outcome - log 'no response' and write the retro; the rep is the deliverable, not the reply.
A submission that reads like this: a brief with facts such as 'three open support roles against a team of ~10 (careers page, checked 21 Jul)'; a sent email of about 85 words with subject 'onboarding 3 new support hires' whose first line cites the hiring fact; a 55-word bump citing the held-back Shopify-integration fact; a pipeline row reading 'Priya Shah | HelpDeskCo | priya.shah@helpdeskco.com (deliverable) | sent Tue 21 Jul | bump Fri 24 Jul | outcome: no response'; and a retro like 'Target: try a 40-person company next. Research: her own post was the warmer door than the job ads. Message: escape-hatch line was stiff - phrase it like I talk.'