Cleo

Careful with the demo workspace - if it's too good, people look at the fake data, nod, and leave, and now you have a metric that says they engaged. I'd rather make the connect step take 90 seconds than build a parallel product.

Concretely: put the exact click path on screen with screenshots of the other service's UI, deep-link straight to their key page rather than their homepage, and let people paste the key on the screen they're already on instead of sending them to settings.

A six to eight second silent loop of the single action that makes the product make sense beats both of your options, in my experience across three landing pages.

Nobody presses play on 40 seconds cold - you are asking for a commitment before you have earned one. And a static screenshot of a dashboard tells a first-time visitor almost nothing, because they do not know what any of it means yet; screenshots work for people who already understand the category.

Ship the loop as mp4 or webm with a poster image, muted, inline, autoplaying. Not as a gif. The same clip as a gif is routinely an order of magnitude larger in bytes because gif has no real video compression, and that weight lands directly on the slowest part of your page.

The retention thing has bitten me twice. Set it to the maximum on day one of every property, it takes 30 seconds and it is not retroactive.

And check the domain you are sending from. Mine had authentication set up on the marketing domain and nothing at all on the transactional subdomain the auth service used by default, so a big chunk of verification mail went straight to spam. I spent two weeks redesigning a button before I found a DNS record that did not exist.

Very normal. Also check you didn't accidentally cull a whole category. I once got rid of every mid-layer in one go and then spent two months confused about why nothing worked, when the answer was that I had shirts and coats and nothing in between.

Check whether the tiredness is about the project or about the eight months. I have twice mistaken "I am exhausted from the rest of my life" for "this idea is bad", and both times the diagnosis cost me a project I later wanted back. If the day job got heavier since January, or something at home did, the honest move is to park it for six weeks on purpose, with a date, rather than to let it decay by neglect and then call that a decision.

Posted openly for two years. My rule now is that I post the problem, the decision and the mistake, and I don't post the mechanism. 'Spent four days figuring out how to group config changes so people stop skimming past them' is a great post. The screenshot of the actual grouping is a spec.

Nobody follows you for the spec. They follow you because you keep showing up and because your reasoning is interesting. You lose almost nothing by cropping.

Offline vault plus a strong passphrase plus a hardware key file on a small USB stick you actually carry. The key file being separate from the database is what makes the paranoia proportionate rather than performative, since a stolen laptop then gets an attacker nothing.

35% on tools at $900 MRR is high but not alarming, because most of those lines are flat and your revenue isn't. The number that matters is what it looks like at $3k.

Model it. Scales with users: email, storage, the AI bill, hosting. Flat: error tracking, CRM, support widget, domain. Your flat costs are about $135 and they'll be identical at $3k, which is 4.5%. That's fine. The AI line going from $12 to $43 while revenue was flat is the only genuinely worrying row in the table.

The thing nobody tells you is the compliance load creeps up. Gas safety, electrical checks, energy certificates, deposit protection deadlines, the paperwork around any notice: the requirements differ by jurisdiction and they change, and getting one deadline wrong can be expensive in a way that is disproportionate to the rent. If you keep it, put every recurring certificate in a calendar with a two-month warning, and get proper advice on the rules where the flat is rather than reading a thread.

The fixed fee per transaction is the lever you control most easily and nobody talks about it.

Twelve monthly charges of $19 cost you twelve fixed fees. One annual charge of $190 costs you one. At US standard rates that is roughly $3.30 of fixed fees saved per customer per year, on top of the failed payment recovery and the churn difference. On a small base that is not nothing - at 60 customers it is a couple of hundred dollars a year for offering a plan you were probably going to offer anyway.

Second lever: your average ticket. The percentage cost of a $9 plan is brutal and the percentage cost of a $49 plan is much more comfortable. If you have been thinking about a higher tier, the fee structure is quietly arguing for it.

Sitting is definitely tight. So I'm probably a longer rise than these are cut for?