← The Marketing Station ~ dish 06 of 10 · the marketing station

Ad Copy Variation

The Chef's habit problem isn't the station can't write ad copy. It's we never hooked the station's output to performance data, so it generates ten mediocre variations forever. Without measurement wired in, every variation is a guess.

The job

You have an offer. A webinar. A consultation. A case study download. You need headlines and body copy for three ad platforms. LinkedIn, Google, email. Ten variations per platform. Thirty variations total. Each one should be a different angle, not just a word swap. One plays on urgency. One plays on curiosity. One plays on social proof. One plays on pain. The angles are different. The conversion likelihood is different.

If you write these yourself, it takes four hours. If the station writes them, it takes twenty minutes. The catch. You have to measure which one actually works. If you don’t track what converts, the station is just spinning wheels. It has no feedback. By month three, it’s still drafting variations in the same pattern it learned on day one. The difference between ship and scrap is Measurement.

The recipe

All seven ingredients still apply. The leverage on this dish is Measurement (Ingredient #6). You have to track what works.

Training matters. The station learns your offer posture. How hard you sell. How much social proof versus urgency. Examples matter. Show the station five ad headlines you wrote that actually converted. Not the ones you liked. The ones that moved. The station learns what worked and builds variations in that shape.

Guardrails matter. Things you never do. You never make a promise you can’t keep. You never use fake scarcity. You never claim social proof that isn’t real. Context matters. What’s the offer. What’s the target audience. What’s the current angle or control if one exists.

How to build it

  1. Define the offer clearly. The thing you’re asking them to do. Download. Register. Call. Schedule. Be specific.
  2. Name the target audience. Who is this offer for. What do they care about.
  3. Pull five ad headlines you’ve run that actually converted. If you don’t have conversion data, pull five that performed better than baseline in your gut. Pull the body copy too.
  4. Note the angle in each. Urgency. Curiosity. Social proof. Pain. Fear. What angle was load-bearing in each one that converted.
  5. Define guardrails. “Never claim a guarantee we can’t deliver.” “Never use fake scarcity.” “Never claim a stat without a source.” 4-5 rules.
  6. Define the measurement success metric. Clicks. Cost per lead. Conversion rate. What signal tells you a variation is working.
  7. Feed offer, audience, examples, guardrails, and success metric to the station. Ask for ten variations per platform. Three platforms. Thirty variations total.
  8. Run them. Track performance for two weeks. Note which angles converted better than control. Then feed that performance data back to the station.
  9. On iteration two, the station knows what worked. It generates new variations in that direction.
  10. By iteration four, you have a control that converts and five variations that beat it on different angles.

What breaks it

  • No measurement defined. The station generates thirty ad headlines. You run them. Nobody tracks performance. By week four, the station is guessing. Output plateaus because feedback loop is dead.
  • Examples are weak. You pull headlines you liked, not headlines that converted. The station learns taste, not performance. Output is creative but doesn’t move.
  • Angles are too similar. The ten variations are all plays on the same angle with different wording. You’re not testing angles. You’re testing synonym quality. You learn nothing.
  • Guardrails are vague. “Don’t oversell.” The station has no idea what that means. It generates copy that sounds authentic to you and hyperbolic to the market.

When it’s working

By week two, you’re running variations and tracking performance. By week four, you have data showing which angles resonate with your audience. By week six, you have a control that converts and you’re using iteration two to beat it. By month two, the station is good at guessing which angle the audience will respond to because it learned from month one’s data.

The signal that the recipe is sharp: you have a clear winner in each ad platform. That variation converts 30% better than the others. The station learned what your audience actually responds to.

Monday Move

Pick one offer. Define the audience. Pull five ad headlines and copy you’ve run in the past year that performed well. Feed the offer, audience, and examples to the station. Ask for ten variations per platform. Three platforms. Thirty variations total. Pick the three that intrigue you most. Run them against your current winner for two weeks. Measure. Then feed the performance data back to the station for variation two. By week six, you know which angles work for your offer.

Dish 6 of 10 on the Marketing Station. Build-note leverage: Measurement (Ingredient #6).