Hacker News new | comments | show | ask | jobs | submit login
Ask HN: Review my startup idea
5 points by doobiedoobiedoo 1470 days ago | hide | past | web | 2 comments | favorite
I'm almost finished building a real user monitoring (RUM) service for live web performance testing.

This is useful for a couple of things: 1) Regression Testing... is your site slower (or better!) than it was before you rolled out a new version or new feature? 2) What page(s) are performing poorly & receiving alerts if not hitting SLA targets. 3) Usage predictions. 4) Analyzing if a performance problem is you or your dns(etc.)

There's no software install required for your server.

I've seen this kind of thing in other products, but it's usually tied to an expensive pro account.

Before going public & writing up the UVP for the landing page I'd like to get some feedback on the idea... would you pay for this, what would you like/dislike having in such a service etc.

Thanks a ton HN.




Hi, user experience and interaction designer here. I especially like talking with early-stage startups because I ask hard questions that save a lot of time and effort in the long run. I also offer this as a service via UX Hours: http://uxhours.com/

1. "Review my startup idea" is your title, but your first line is "I'm almost finished building." What if everyone says your idea is terrible and points out hundreds of competitors and software and reasons why yet-another-one-of-these is a bad idea? How will you feel? Will all of that time you spent be wasted? Will you be able to be competitive?

2. You might not get honest feedback because you're so far along. Normal people (especially friends and family) don't like to be discouraging, whereas designers do it for a living. You might also get honest feedback, but ignore it, because it conflicts with what you've already built, and people naturally try to preserve their own egos and justify their own efforts.

I'm not a performance engineer, but when I worked somewhere that had one (these numbers refer to your points):

1. There's no way we'd outsource regression testing. We had three load-testing environments. First, the developer's local machine could be configured to do test timings of whatever they were working on before they started, and then after they finished, to do a basic gut-check. Two, we had a partial internal mirror of production (ten servers, maybe?) and a dedicated person and load-testing software. This was used after QA, before final sign-off and deployment. Three, we could slow down production and actually run it on the production machines by taking some of them out of rotation. This was done when we needed to be really sure about something. We would never expose untested software to real users.

2. Our "pages" supported partial failures: we had a cloud CDN providing anonymous access, our servers for logged-in users, and pages would still work with reduced functionality even if certain back-end services were down. "SLA targets" hits a whole raft of dependent services that we wouldn't give you access to.

3. Automated usage predictions are useless for new features. That's what customer research is for.

4. We had an entire operations team to tell us that.

5. Cost was not an object because this wasn't "the web site." It was "The Business."

6. When we had trouble, it wasn't about the cost of the software, or even the usability of the software. It was not having a person who could not just do the monitoring, but also diagnose the problem and help us fix it. Maybe they were sick. Maybe the on-call person didn't have the right expertise. Maybe the ops guys called the wrong person.

Turning actual user behavior into load test data might be very useful if you're a one person team, but it can't replace customer research and a testing team and a load testing team, so positioning is important.

Friends who wanted this sort of software were concerned with:

1. Recording behaviors and automatically aggregating common repeated actions into sets of flows.

2. Ways to automatically turn that data into tasks that can be run in load-testing software. The easier this is, the better. If I have to also learn how to use load testing software, you suddenly lose a lot of value versus writing something myself.

3. Proxying for Google Analytics. I already send 100% of my usage data to Google. They already hook everything that happens. Why do I have to install something else in my pages that will do the same thing? That slows down load times and might introduce conflicts when you get a click and they don't, or vice versa. Why can't you just piggyback on those events? Or send the events on to them? Or get the events from them?

In short, they didn't want a tool, they wanted "install this one line of JS and next week we can run statistically valid load tests with one button click for $N a month."

(And going back to my first paragraph, this is probably what you'd get out of an hour conversation.)


Wow, that's feedback!

1&2) I'll definitely be doing that with the next one.

-- 1) I wouldn't expect complete use QA to be exposed, just seeing the real speed numbers of browsers. 2) what kind of services? 3) sorry, I meant traffic patterns... busy weekends etc. 4) ok 5) ok 6) agreed, positioning is everything -- 1,2,3) again this is mostly a speed performance thing... as for GAnalytics my own experience is that, for speed, the sampling of data is quite low & the usefulness of the reporting equally so. This would be using insertable, nonbocking js code.

Thanks for the awesome feedback!




Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | DMCA | Apply to YC | Contact

Search: