Cost by channel
Experiments
Each test had a hypothesis, a budget cap, a primary metric, a guardrail and a window, written before it ran. The decision comes only from those rules, checked in this order: most of the cap spent with no held demo, guardrail, no wins yet, too few wins, stop bar, scale bar.
What changed this month
Held demos, September against August, split by channel and practice type, so the few cells that drove the change are found before anyone decides what it means. Credit follows the setting above.
Rows a person should check
These rows are counted by a stated rule, but a rule can be wrong about them. Each one says how it was counted and why it's worth a look.
Assumptions
- Every office, group, date and dollar amount is made up: the lead rows, and the spend for April to September 2026.
- A paid practice is one office location that started paying by September 30. A group that pays for only some of its offices counts only those; the rest are "expansion pending".
- A signed account is one customer, whether it's one office or a group of 25.
- Cost is media or vendor spend plus field costs (booth, travel, food). Founder hours count only when "Include founder time" is on, at the rate you set.
- Click (UTM) credits the ad or link the booking came from. Self-reported credits the answer to "How did you hear about us?". Blend takes the answer when it names a channel, otherwise the click. Anything else is Unknown and credited to no channel.
- Each test's leads are tagged to that test when they arrive (a partner code, a booth scan, a postcard code or an ad campaign), so test decisions don't change with the credit setting. The channel table does.
- If the same office books twice, it counts once, at its earliest touch, and keeps the furthest step either booking reached.
- Under 3 paid practices (or accounts, when counting accounts) is "too few to call". One lucky deal isn't a channel.
- The scale and stop bars, minimum wins, guardrails, caps and windows for each test are assumptions, as is the $100 an hour founder rate.
- Recent spend has had less time to turn into paying practices (a pilot runs about four weeks), so the newest months look more expensive than they'll end up.
- Money is kept in whole cents and shown in whole dollars, rounded down.
Questions I'd ask first
- When a group signs, do you count one customer or one per office, and does price follow offices or providers?
- When the demo booking's source and the "How did you hear about us?" answer disagree, which do you trust?
- Do founder hours at lunch-and-learns and conferences count as acquisition cost?
- What payback period would make you put more money into a channel?