SimplyCodes is our coupon site, and its whole claim is that its codes work when other sites’ codes do not. That claim rests on three numbers, sometimes called “coverage” or “accuracy”:

  1. Machine coverage: the share of verification work the machine finishes without a person.
  2. Code success rate: the share of codes we publish that work at checkout.
  3. Rival coverage: of the working codes that rival coupon sites serve, how many do we also serve. Rival sites here means sites such as RetailMeNot, Honey, and Coupons.com; the written method names the full list.

The first two are measured and have targets. The third has a written method and a frozen panel of about 500 merchants, but nobody has run it. This mission is the first reading of the third number.

Our goals for this mission

Until the rival number exists, half of our claim is unproven. We can say our codes work. We cannot yet say how our coverage compares with the rival sites. A first credible reading of that number matters more than a small gain on a number we already track, because every later decision builds on it.

What you would do

  • Kickoff. We hand you four things: the written method, the frozen panel of about 500 merchants, the current readings and targets for the other two numbers, and a file of our own verification verdicts for the codes on the panel. You and the panel's owner agree the sample, the time window, and how the error bars are computed, before you collect anything.
  • Collection. For each merchant in the sample you collect the codes the rival sites list, by hand from their public pages, and record where and when you found each one. You match them against the verdict file to find the rival codes our verification confirms work, and you check whether SimplyCodes also serves each of those.
  • The reading. You compute the reading with its error bars, and you report the share of rival codes our verdict file could not grade as its own line.
  • The write-up. You write the method page anyone could re-run on the same files, and one page on what you found and what you left alone.

What success looks like

  • The first rival coverage reading exists, dated, with a method anyone can re-run and error bars stated beside it.
  • The codes and pages you collected survive as a file, so the next reading can be compared with this one.
  • A small or unflattering number is reported as it is, with an account of why.
  • The share of rival codes you could not grade is stated, never folded into the reading.

Ground rules

  • No access to our systems. Inputs arrive as files; outputs return as files.
  • Codes are collected by hand from public pages. No scripted or automated collection.
  • You do not test codes at checkout. Which codes work comes from our verdict file; your job is the comparison.
  • The method page states the sample, the window, and every exclusion. A reading without them does not count.
  • Scope, timing, and pay are set in the contractor agreement, at your stated rate.

What happens after you send a proposal

3 steps, and a person reads every proposal.

  1. You send a short written proposal. It covers how you understand this mission and how you would approach it, and you attach your resume or a LinkedIn profile.
  2. We read it and reach out with questions. There is no test and no exercise.
  3. If there’s a fit, the mission begins under a contractor agreement and an NDA, paid at your stated rate, with kickoff in Santa Monica.