I built ecommerce analytics that prices each fix in yen. Product's done, 0 customers — here's where I am.
I'm a solo developer in Japan. I run a small online store myself, and for years my analytics was "installed but never opened" — I'd glance at the pageview count and close the tab.
So I built the thing I wanted: NextRise Analytics — ecommerce analytics that doesn't stop at showing you numbers.
What it does
- One line of script. No cookies, no localStorage, so no consent banner.
- One screen built around Revenue = Sessions × CVR × AOV, split the way a store owner actually thinks: source, device, product, landing page, day × hour.
- Every morning it looks for gaps in your own data and writes them as cards with a number attached. From the public demo store: "Mobile shoppers convert at 4.4% vs 14.0% on desktop → these pages are slow → about ¥318k/month if you recover 30% of the gap." One formula, one assumption you can change, and the math is on a public page. If a card can't be backed by a formula, it doesn't get a number.
- Four weeks later it re-measures and tells you whether the fix actually worked.
- Orders come straight from the cart (WordPress — WooCommerce and Welcart — today, more carts landing as their reviews clear), so revenue is the store's own number, not an estimate from the tag.
Where I am
- Product is finished and live, in Japanese and English. $49/mo, 7-day trial, no card.
- The WordPress plugin went live in the wordpress.org directory a few days ago, and started showing up in the plugin search inside wp-admin a couple of days after that. That's the first channel where a store can find me instead of me finding them.
- Two app-store reviews sitting in a queue — one is a 3–6 week wait, and it bounced me once because the reviewer could reach my own checkout page from inside the app. Nothing to do but wait.
- 0 external customers. Marketing started less than a week ago: no ads, one long-form post a week, and being useful in places like this.
- Biggest lesson so far: the advice engine was maybe 10% of the work. The other 90% was making the numbers agree — order count vs conversion count, revenue definitions, timezone boundaries, tax. If the number on the card doesn't match what the store's own admin says, nobody trusts the card, and you don't get a second chance at that.
- Most surprising thing: I opened 114 major Japanese ecommerce sites in a real browser last week. 96% run GA4 — and 46% still carry a dead Universal Analytics tag, two years after Google deleted the data behind it. "Installed and forgotten" is the norm, not the exception. Write-up plus the anonymised 114-row CSV: https://nextriseanalytics.com/en/blog/japan-ec-114-sites-survey/?utm_source=indiehackers&utm_medium=social
What I'd love from you
- If you run a store: does a card like that make you want to act, or does the money figure make you suspicious? I went back and forth on whether to show money at all.
- If you've sold analytics — or any "here's what to do next" tool — what actually got you your first 10 paying customers?
Demo, no signup: https://app.nextriseanalytics.com/demo
Site: https://nextriseanalytics.com/en/?utm_source=indiehackers&utm_medium=social
Happy to go into the cookieless side, or into how the cards get their numbers.
showing the math is what makes the number believable, without it i'd be suspicious too. and honestly the survey post you did probably does more for first customers than any outbound would, that's the kind of thing that gets shared on its own
Thanks — that's the reason the estimate page exists. Every card links to how its number was built, and if I can't write the formula, the card ships without a number. It made the cards less punchy, and I think more usable.
You may be right about the survey. It took three days and has already started more conversations than everything else I've done combined, including this post. The uncomfortable part is that it says nothing about my product — which is probably exactly why it travels.
Really like the 'no formula, no number' rule - most analytics tools print numbers nobody can defend. The 4-week re-measure turns reporting into a loop; yen-per-fix is what gets a store owner to open it daily. Building analytics ourselves (amami.dev), same lesson: conclusions must arrive where the owner already works, or the dashboard stays closed.
"Conclusions must arrive where the owner already works" is the line I needed to hear today. The delivery side is built — cards are generated nightly and there's a mail pipeline running — but a week ago I deliberately postponed the weekly digest email on the grounds that it was a distribution problem, not a product one. Your framing says that's backwards: a dashboard nobody opens isn't a product either. I'm going to revisit it.
The "numbers nobody can defend" part is what cost me the most time, for what it's worth — not the detection, but making my numbers agree with the store's own admin screen, order for order. Good luck with amami.dev. If you ever want to compare notes on the boring layer (order count vs conversion count, tax, timezone boundaries), I'm up for it.
The first-customer measurement: the ¥318k card gets seen, but does the owner run the fix and come back four weeks later to see the re-measured number? That's the signal that separates "interesting insight" from "worth paying for." Most analytics tools die because people see the data but can't measure whether the action worked, so they never trust the next recommendation. If you can show someone that the card's number went down after they fixed it, you've proven the data is real, and you'll own that metric forever. That's the wedge for first ten.
That's exactly the bet, and it's the one part I built before I had anyone to sell to. A card that survives three consecutive nightly sweeps opens a trail row: the card's own numerator and denominator at first sight (deliberately not the money figure — that moves when you change the recovery-rate assumption), the date, and an optional "I acted on this" button. Four weeks later it re-measures baseline-28d vs latest-28d with the same two-sample test the card itself used, and labels it improved / unchanged / worsened / inconclusive.
Two rules I forced on myself there:
worsened is shown, in the same place as improved. A tool that only reports its own wins has no evidence value.
the wording never claims causation: "18% → 24% (+6.0pt, bigger than chance)" plus a line saying other things also changed in those four weeks. I'd rather be believed than impressive.
Honest status: recording started Sept 1, the first verdicts land in early October, and they'll be on my own stores. So I can't answer "did an owner act and come back" with data yet. The day I can, that's the number I'll publish here.
The ¥318k/month figure is the interesting wedge. Have any store owners actually acted on a card and measured the result, or is proving that the number changes behavior still the main unknown?
Still the main unknown, and I'd rather say that plainly. Nobody outside me has acted on a card yet — 0 external customers, so the honest answer is "not proven."
What does exist is the measurement: cards open a trail with their raw numerator/denominator, get re-measured 28 days later, and get labelled improved / unchanged / worsened / inconclusive — worsened included. Recording started Sept 1, so the first verdicts are early October, on my own stores.
My working assumption is that the money figure isn't what makes someone act — it's what makes them pick that card over the other nine. Whether the fix actually happens seems to depend on whether it's small enough to do that afternoon, which is why the cards say "compress the images on these 3 pages" rather than "improve mobile UX." But that's a belief, not a result.
That’s a useful measurement setup, even if external behavior isn’t proven yet. If you’re open to it, what’s the best email to reach you on?
Sure — contact@nextriseanalytics.com reaches me. Happy to keep it in the thread too if it's something other people here would get something out of reading.
Thanks! I’ve just sent it over.
Looking forward to hearing your thoughts whenever you have a chance.
For first 10, I’d lean into stores where you can point at one concrete money leak from their own site, then offer to walk through the demo with their numbers. Also worth testing low-friction launch surfaces where founders can list a product without a huge audience. I’m playing with a free one built as a founder challenge: 8 events, one all-round score, top-five SaaS links: https://www.buildersbenchmark.com/
That matches where I landed. The opening move is one concrete leak from their own site, not a feature list — I have a public URL grader (speed, which analytics tags are actually installed, structured data) and the plan is to run it on a store, send the single thing that's costing them the most, and offer to walk through the demo with their numbers if they want it. First batch goes out this week; I'll report the reply rate here either way.
Thanks for the pointer, I'll take a look at Builders Benchmark. Low-friction surfaces are what I'm shortest on — my ICP is Japanese store owners, who aren't on the usual launch sites.