How Skylight grows

Non-subscribers on the Skylight device could scan a QR code to subscribe; it used to open a web checkout. Flipped it to Apple in-app purchase so the user scans and subscribes on their phone in one step. Skylight also kept IAP rather than moving subscriptions to the web after anti-steering changes, because their test showed the conversion gain from less friction outweighs the 30%/15% Apple fee (only a fraction of subscribers pay via Apple, since most subscribe in the owned web channel).

Plus subscription had been $39/year (same SKU for Frame and Calendar). Asked what a Calendar-only company would charge, then ran a batch of variants varying both the subscription price and the hardware discount when bought together (tracking hardware conversion, margin per unit and subscription attach). $99 maximized revenue slightly in the spreadsheet, but surveys, one-on-one conversations and focus groups showed $99 was near 'disgust / not worth it yet' territory, so they chose $79 as a middle ground they were confident in, keeping room to go up later as more value is added. Applied to new customers only (see t2).

Skylight has never offered a free trial; most people subscribe at purchase. Ran a free trial test in two variants: one taking the customer's credit card and one without. Run as a side project with ~5% of the roadmap, no follow-up iterations.

Regularly poll customers: is it life-changing, very awesome but not life-changing, or worse? Hypothesis: the 'life-changing' group drives most word of mouth (each bringing many new customers). The CPO's explicit goal is to build features that move this to 50%.

On the owned channel (myskylight.com) the subscription is turned on automatically at purchase and buyers get a discount on the hardware if they sign up with the subscription, plus a free month and the option to opt out. Most people don't opt out. Not possible on Amazon or in retail stores. The size of the hardware discount was also a variable in the price tests (t1).

Hardware margin means each order recovers its marketing cost immediately, so there is no payback period and no need to raise equity to fund growth. Advice for subscription apps: turn a ~$5/month plan into a $60-$70 upfront annual payment to get the same effect on CPA and payback. Marketing spend is capped where it still pays back (e.g. ~$130 CAC) rather than pushed to contribution zero.

Kept a weekly cash flow model early on, since inventory must be bought weeks ahead of Black Friday and COGS are ~50%. Grew 'fast, but not too fast' so lenders stay comfortable, and found lenders more risk-friendly than banks (which balk at 4x growth plans and unsellable connected-device inventory). Paid 15-20% annualized at times; traded gross margin for keeping ownership.

Rule: what the customer bought the $300 device for stays free (sending photos to Grandma on Frame; events, tasks and chores on Calendar). New, easy-to-grasp modules go behind the paywall from launch (videos on Frame; meal planning, recipes, AI features on Calendar), often without testing, since moving a feature out of the paywall is easier than in. Photo screensaver on Calendar was paywalled at $39/year from launch. They watch reviews and interviews for anger or 'disgust' and will pull features out when sentiment turns, even if the paywall lifts numbers.

Inside the product, premium (Plus) features are not called out as premium for subscribers, so at renewal they don't tally which bundle features they used. Related finding: changing how the bundle's contents are merchandised to buyers rarely moved the attach rate.

Calendar shipped November 2018 as a quick launch; the deep software was hard to build and under-resourced. In 2021-22 they 'gunned marketing' on it. It failed, and they considered killing Calendar. Lesson: until the product is around 40 NPS, marketing spend gives a false negative because you can't tell whether marketing or product is the problem. After investing in the product, Calendar took off in fall 2022.

Multi-channel distribution: direct on myskylight.com (controls customer relationship, auto-on subscription), Amazon, and retailers including Costco and Best Buy, plus the same abroad. Retailers generally take only one player per category, which works as a moat. Trade-off: in Amazon and retail the subscription can't be bundled at checkout, so attach is lower and the product must sell the subscription after setup.

Tried several changes to how the Plus bundle's contents are merchandised to buyers at purchase/setup.

Did not announce the $39 to $79 increase in advance and did not apply it to existing subscribers; they were moved to a 'Legacy Plus' plan (Frame and Calendar access at the old price). New price only showed to new customers. Policy: anyone who had been considering a purchase and reached out asking for the old $39 price got it honored, even a year later.