Your First Two Weeks

The failure pattern for a new speaking tool is almost always the same: it gets introduced to students in the same week the instructor is still working out how it behaves. Both sides are confused at once, the first activity goes badly, and the class decides it doesn't work. Two weeks of deliberate sequencing avoids nearly all of it.

Lemon says: Learn it before they do. That one rule prevents most of what goes wrong.

Don't build a unit. Build one activity, for one thing your students genuinely need to practice, and put real care into the scenario. Everything you learn here transfers to the next twenty.

• Copy something close from the Library, or use Create with AI with a rich prompt
• Rewrite the scene until it has a complication that sustains the full session
• Rewrite the character until it has a name, a manner of speaking, and a goal
• Set the level one notch below your instinct
• If your Dashboard offers Request Free Student Access, send the request now — it needs approval before students can practice

Lemon says: One good activity teaches you more than five rushed ones.

Use the Try button and run your own activity at least twice. Once as a cooperative student, once as a struggling one — hesitant, short answers, an English word or two, a deliberate mistake.

• Does it fill the minimum time, or run dry early?
• Does the character hold its role when you go off-topic?
• Is the register right for your course?
• Does the feedback report say anything useful?

Lemon says: The struggling-student run is the one that finds problems. Do that one twice if you only have time for two.

Most first-week friction is technical rather than pedagogical: microphone permissions, browser choice, finding somewhere quiet enough to speak. Handle it before it collides with a deadline.

• Send the setup instructions ahead of the first activity
• Tell students explicitly that they can practice more than once
• Tell them what to do when they get stuck mid-conversation
• Add a syllabus line so expectations are on the record

Lemon says: The "Student Onboarding Kit" guide has text you can copy straight into an email or syllabus.

Make the first real activity ungraded. The goal is that every student successfully completes one conversation — not that anyone demonstrates mastery. Save rubrics for the second or third activity.

• Assignment type: Submission required, with Ungraded review — completion-only work keeps transcripts private, and you'll want to read some
• Generous availability window, no tight deadline
• A gentle scenario with plenty of scaffolding

Lemon says: First activity graded is a mistake I see a lot. Let them fumble it once with nothing riding on it.

Two days after the activity opens, look at Student Work. You're not grading — you're checking whether the mechanics worked and whether the activity behaved the way it did in your test.

• How many students completed it? Low completion means a setup problem, not a motivation problem
• Open three transcripts — one strong, one middling, one weak
• Did sessions end the instant the minimum was reached? That means the scenario ran dry
• Did anyone get stuck in a way your scaffolding settings should have caught?

Lemon says: Three transcripts is enough. You'll spot the pattern immediately.

Make your changes while the evidence is fresh, then plan the rest of the unit. By this point you know how your students behave with the format, which is the knowledge that makes everything after this fast.

• Fix the one thing that went most wrong — usually scenario depth or level
• Build the next two activities using what you learned
• Introduce a rubric now if you plan to grade
• Reference the feedback in class so students know it's read

Lemon says: Mentioning one student's feedback out loud in class does more for engagement than any feature I could build.