The length debate gets a lot easier once you assign each format a job. In a lead ad, a 6–10 second clip has exactly one assignment: stop the thumb and make one promise. It will not close anyone — the closing happens on the landing page and in how fast you call back. So judge short clips on click quality, not on whether they "explain" your service. The longer cut is where trust gets built, but most of that trust-building can live on the page the ad sends people to, which is cheaper to produce once than reshooting ads.
On being on camera: yes, a real face outperforms faceless footage for home services, but it doesn't need to be a performance. The low-budget version that works is a tech or cleaner saying one plain line to the phone — "This kitchen took us about three hours" — over actual job footage. Fifteen seconds of real beats a polished minute of stock.
Before/afters: keep them honest or don't use them. Same angle, same lighting, no filters, and get the customer's OK before filming inside their home. Platforms aside, one "that doesn't look like the same room" comment under your ad undoes the whole campaign in your own town.
And the unglamorous one that decides whether any of this is measurable: set the tracking up before you test lengths. A different phone number (or at minimum a "how did you hear about us?" field) per ad set. Attribution is a big part of my work at LocalCMO, and the most common thing I see is owners A/B testing creative for a month, then realizing the leads all landed in one undifferentiated pile — so nothing was actually tested. Nail that first, and the short-vs-long answer will show up in your own numbers within a few weeks.
