Hacker News new | past | comments | ask | show | jobs | submit login
Web Services Our Startup Relies On Every Day (mygengo.com)
198 points by robert_mygengo on Jan 18, 2011 | hide | past | web | favorite | 66 comments

I'll second that Xero is an excellent service and it really doesn't get as much press as it deserves in the states. They're very easy to use and have done an heroic job building out both the software and support.

HOWEVER, their entry-pricing is incredibly irritating. For $228/year you can't add any more than 5 invoices or reconcile more than 20 lines of bank statement per month. For the entire price of quickbooks every single year you should be able to at least use your accounting system.

But for only $29/month you can. How much is your time worth? If the service saves you even an hour per month, surely it is worth it.

>How much is your time worth?

I hate this argument. It's so narrowly focused (and for some reason, often regurgitated). Money adds up a lot quicker than that one hour of your time. If you save yourself 10 hours of work per month because of a couple services you pay for, but end up spending an extra $300/month while your startup isn't making any money -- is that actually smart time/money management?

There are more factors involved than simply how you value your time. The stage your business is in, how much money you're pulling in, how many other things you need to get done per day, is your startup bootstrapped, do you still have a day job, etc.

I think the question still applies. How much your time is worth does not need to be hypothetical - it can take into consideration how much your employer (presumably your startup) can afford to pay you for your time.

two words - "opportunity cost".

Why do you think that opportunity cost is specifically applicable to this situation? If it's "just in general" then I don't believe that opportunity cost is a good argument for continuing the meme of "How much is your time worth?" in these kinds of discussions. If it is applicable, it would have to be on a case-by-case basis to even make sense, no?

A price should either be "cheap and cheerful" or "real and worthwhile". $60/year for a single company is cheap and cheerful, $228/year is very much "Real and Worthwhile".

$228/year may be reasonable for a "limited" piece of software but it is too much for crippled software. Make it $60, make it work or take it away, it's an unnecessarily irritating package.

I've recently consulted on a Xero API gig and trust me their API and contributed code is totally awesome, too.

I'm not convinced about the need for SendGrid. Learning to set up an email server isn't that hard in this day and age, and we seem to get very decent deliverability with our own SMTP server - and don't pay a penny for any emails sent.

I'm planning to build an email subscription feature for swombat.com soon, and was looking at options like SendGrid, but I just can't justify the cost. When you send people a sign-up confirmation email, they expect it right now. If it's not there in a couple of minutes, in my experience, even the least tech-savvy have learned that they need to check their spam folder.

I was talking to the guys from mailgun.net recently, which provides a similar service to SendGrid, but also makes it easy to receive e-mail via webhooks. They told me a ton of interesting things about how this stuff actually works.

Getting high deliverability is actually incredibly complicated. You need a dedicated IP address to send mail from, which you need to be whitelisted by the big ESPs (E-mail Service Providers). To get it whitelisted, you need to "warm it up" before you start using it properly. You need to be very careful about setting the right headers. You need reverse DNS configured correctly, you need SPF and DomainKeys set up, you need to make sure the IP/domain you are sending from hasn't previously been used for spam... and even after all that, you probably need to set up a relationship with AOL/Gmail/Yahoo etc to make sure your stuff gets through. You also need to watch for their bounce / rate limit notifications and dial back your sending rates to match.

There's a whole lot more to it than just setting up an SMTP server.

All of this apart from the relationships with ESPs is relatively easy to do if you're a competent sysadmin.

As far as the ESPs are concerned, all of them appear to accept our emails just fine, except for Hotmail. We've largely given up on Hotmail, but then everyone who uses hotmail seems to be aware that it marks everything as spam and so they check their spam box as often as their inbox, from the looks of it.

It's a nice convincing argument. It just hasn't borne out in practice for us. Email is critical to us, and it gets delivered just fine the vast majority of the time. We've had about 2 cases in 3 years where we needed to intervene manually to talk to some kind of ESP to get us unblocked, because they'd randomly decided to blacklist our IP. That's not worth $80/m.

The trouble with hotmail is they sometimes just blackhole what they classify to be spam, instead of filtering it into the Junk folder. It doesn't even bounce, it just disappears.

Gmail does this as well. Our outbound mail to Gmail accounts was just disappearing until we fixed a reverse DNS record.

Really? I've come across loads of people who have had the disappearing email problem with hotmail (including myself on a couple of systems), but this is the first time I've heared of anyone having it with Gmail.

Getting high deliverability is actually incredibly complicated.

Urban legend.

You need a dedicated IP address to send mail from, which you need to be whitelisted by the big ESPs

That is nonsense. Nobody reaches out to the ESPs to have their IPs "whitelisted" in advance. You reach out when, if and after you've been blacklisted, in order to get unlisted.

To get it whitelisted, you need to "warm it up" before you start using it properly.


You need to be very careful about setting the right headers. You need reverse DNS configured correctly, you need SPF and DomainKeys set up

Yes, there's a few things to watch out for, but it's all pretty well documented nowadays and you only have to do it once. SPF is pretty easy with http://old.openspf.org/wizard.html

you need to make sure the IP/domain you are sending from hasn't previously been used for spam

A quick check at a few of the largest blacklists (spamhaus et al) takes all of 5 minutes. Another 5 minutes that you have to spend exactly once.

[..further marketing pitch for MTA SaaS snipped..]

Well, to sum this up.

It seems many people have a strong misconception of what SendGrid, MailGun and friends actually do. You don't magically get higher deliverability only because your mail is passing through them. They don't have mythical "whitelist" slots at the ESPs. They're not even on the radar of the ESPs.

What you get from SendGrid et al is a (hopefully) properly configured MTA and DNS, to push your mail through. From there your outbound mail is suspect to all the usual blacklisting and scrutiny at the receiving endpoints. And that's all - no black magic at work here.

> That is nonsense. Nobody reaches out to the ESPs to have their IPs "whitelisted" in advance. You reach out when, if and after you've been blacklisted, in order to get unlisted.

That's nonsense, lots of people do that. Only you don't usually go direct to the service provider, see following links:




That's nonsense, lots of people do that.

Lots of people also believe in astrology. Anyone is free to do what they deem good for their karma, you just shouldn't expect a measurable impact towards your deliverability. The big ESPs don't look at these lists.

> The big ESPs don't look at these lists.

Some of them do.

AOL uses GoodMailSystems.



Hotmail uses ReturnPath.


Yahoo seems to run their own. I remember them using one of the big ones but I can't find any mention in their help pages now.


Google only asks for best practices, but it's possibly they look at some of the whitelists for more data.


Oh my, do I have to put it even more bluntly to get the point across: These commercial whitelists are effectively extortion scams and affiliate rings.

Their sole purpose is to extract money from you. They're businesses, that is their business model.

The value they provide in return is rather minimal - unless getting e-mail out is at the very core of your business and you need to pull every last string to absolutely maximize it.

Otherwise there are cheaper ways to get reasonable deliverability even at AOL and Hotmail. The default deliverability is already quite good, which is what services like SendGrid largely bargain on. I firmly doubt SendGrid signs each of their customers up for the aforementioned whitelists. They probably offer "assistance" with the process, which probably amounts to sending you a link to a nice HowTo document, and which probably yields them a small affiliate kickback when you actually follow through.

> Oh my, do I have to put it even more bluntly to get the point across: These commercial whitelists are effectively extortion scams and affiliate rings.

You made a few points.

First is that nobody reaches out to get themselves whitelisted in advance. Lots of people do that and some ESPs even offer their own whitelist request forms.

Second is that none of the big ESPs look at third-party whitelists. AOL and Hotmail say they do.

Third is that even if the above is true, the impact on deliverability is minimal. I would guess you are right about that but I'd love to look at some data.

Intermediaries like SendGrid would be well within reason to be paying for such services. I don't think they have to signup their individual customers on the whitelists. They just say they are intermediaries and promise to vet their customers.

I think anyone sending low volumes of email is much better off using an intermediary. They get to use well-configured mail servers, IPs with good reputation, etc. Doing it yourself, IMHO, is not worth all the trouble compared to paying $20/month.

>>> "You don't magically get higher deliverability only because your mail is passing through them. They don't have mythical "whitelist" slots at the ESPs. They're not even on the radar of the ESPs."

That's absolutely right, you don't. But email deliverability turned from a massive mystery and headache for us, into a sure-fire deal, the day after we implemented SendGrid. We develop translation/localization related apps, not email apps. So for us that's cool :)

If they're a good deal for you then there's no arguing with that.

I didn't mean to say they have no value proposition at all; the API and stats they provide can sure be useful for some.

All I was saying is that their raw deliverability is no different to running your own server. And your own server doesn't charge a few pennies per mail...

Sure, I think it's a valid point. No beef there.

I also think it's a good idea to get two separate IPs. One for signup/notification emails, and a separate one for marketting. People will report marketting emails as spam, even if they've signed up for it.

Not sure how many emails you send a day, but I signed up for sendgrid almost immediately after I learned it existed.

ISPs have lots of arbitrary policies. For periods of time my emails have been blocked from AOL and AT&T accounts, which make for a lot of headaches and support emails to deal with. (I'm just sending out forum registration, activation and thread subscription emails. no mass mailings.) For a while, we just told people to use non-AOL addresses.

Also, I haven't noticed a delay in email deliverability. It seems to deliver quickly.

Edit: If it matters, for reference, my forum sends out 6000-7000 emails a day.

I use SendGrid, too. It was a no-brainer: $80 a month is cheap, I am expensive. I was quite expensive indeed dealing with persistent minor clusterflops like, e.g., automated forum spambots for the forum I host for college debaters (an old hobby) caused the BCC mailserver they were freeloading off of to have enough of a spike that a large multinational advertising company decided BCC "Thanks for paying me $30 Cindy, here's the software you just bought" emails needed to be consigned to the seventh circle of hell sandwiched between Viagra spam and their own customer support requests.

For us it's just another thing that's not our core competency which we're happy to have outsourced so that other people worry about security/maintenance etc... Similar rationale for using Spreedly [1] for subscription billing.

In the future it may be the case that we move that in house, but right now it's working out really well.

We're using PostMark [2] which amounts to $1.50 per 1000 emails, which is cheap enough for our uses. It's one less thing to worry about so we can focus on more important tasks such as finishing the v2 we're about to launch!

  [1] - http://spreedly.com/
  [2] - http://postmarkapp.com

you get decent deliverability now - what are you going to do once that stops happening?

one of the points of SendGrid is that you pay them to worry about that, so you can concentrate on your core competency.

one of the points of SendGrid is that you pay them to worry about that

Where are you getting that from?

I see nothing on the SendGrid page to suggest they track your blacklist status or feel obliged to help with unlisting when it happens.

We use Postmark, (we've sent a few million emails through them) and they indeed do this for us.

I don't think you're wrong — I just think that for $80, we get a lot of peace of mind once we were sending thousands of mails per day. Particularly because we're sending job notifications to translators too — so if they don't see them, they might not think to check spam.

I dropped the idea of using Postmark after I read this blog-post: http://blog.postmarkapp.com/post/1431120311/oddities-in-gmai...

In there they admit that mails sent through Postmark still end up in Spam-folders or black holes. That was basically the only added value I'd get from using them.

This isn't a technical question, it's an economics question. If you would rather put in effort than dollars, then do it yourself, and it will work great for a while. But if your business takes off, you will accumulate email-related problems like bacteria, and suddenly that $20/month starts to sound reeal worth it.

My company uses it because we run our web front ends on various cloud providers, most of whose IP space is blacklisted. I don't even bother to verify that they are anymore and just assume that they are. Sendgrid gets rid of this entirely, which is a pretty big win.

As usual, the only people who make money on the gold rush are the shovel sellers

LOL. I guess we make shovels too though. So no biggie.

Wow. Looks like you have outsourced every imaginable job when there's a webservice for it. Are you doing this because of bad experiences when servicing your own or did you do this from the start? Did you compare the running costs/reliability between outsourcing and self administration with a positive result into outsourcing?

Interesting. I think most of this stuff is not a core competency for a startup.

I guess my responses are:

1. Zendesk We needed a good system, not really outsourcing of support. So, we do our support in-house but we use the Zendesk SaaS system to manage it.

2. GetSatisfaction As above.

3. SendGrid Email deliverability was not a piece of expertise we had in-house, especially when we were 2-3 people. Now we're 10-ish, our sysadmin guys STILL don't want to do this stuff.

4. MailChimp Hand-coding HTML emails more than once was boring.

5. Apigee This is something we were on the fence about. The metrics visualization probably swung it.

6. String No-brainer :)

7. ExpressionEngine Still a bit of friction in the decision between self-coding or using EE. But it's the 'annoying' stuff that swings it like plugin availability for new functionality, rather than do-it-yourself.

8. Chartbeat Couldn't do this ourselves.

9. Mint Could do this ourselves, too cheap to bother.

10. PivotalTracker As above.

11. Salesforce Industry standard, makes sense to outsiders, reliable.

12. RightSignature Awesome. No point building ourselves. Legal worries too if we did.

13. Gotomeeting Yeah no way we want to build this.

14. Xero The non-enterprise stuff we do is billed through our own app. But Xero makes sense for invoicing and proper accounts. Risk is, if you build it yourself, you need to know GAAP inside-and-out AND still it's a concern for outside auditors I think.

15. Dropbox Yeah no way.

BONUS: Office Glico Well, we could get an intern to go to the mom&pop store every week. But no.

Thanks for this long version! Might learn to stop trying to maintain every solution on myself.

Wow, having used EE, I always thought it was cumbersome to use, prone to running out of memory very easy, and did not offer much over regular php or Drupal/Joomla.

>Salesforce is basically just a very flexible database with hundreds of custom fields, i.e. to a programmer it seems like a $25/user version of phpmyadmin... but it’s money well spent.

That is the most underwhelming endorsement of salesforce.com I've ever seen.

+1 for apigee - we just migrated our web services and the stats, graphing and real-time debugger are excellent. Migration was super easy.

This looks really cool. I'm glad I found it!

I use a few in that list too, but in particular SendGrid and DropBox have really been life-savers (paid user of both).

They both solve a problem that is otherwise quite complicated and boring to solve, in a very simple, hassle-free way.

Seems like SendGrid caused the vast majority of comments so far. I feel like Xero, Apigee and RightSignature should get more love. What do you guys think?

I'll add to the list my favorite service I don't ever hear talked about. SpamStopsHere:


The killer feature for me is a very low false positive rate. Very few legitimate emails get blocked, based on the way they are filtered.

We use RightSignature to handle the signing of all internal/external documents. I'm biased, but I think it's the best SaaS document signing application available :)

I prefer uservoice to getsatisfaction, it doesn't have as many features, but it's easier to use. I think that's more important for a social support tool

Does anyone recommend any SendGrid alternatives? They look good, I'm just wondering if anyone has had success with another provider also.

I'm using critsend.com - it's just "pay as you go", not a monthly fee. $1/1000, with discounts as you go up in credit purchases. I've had no problems with them in the last year or so. No - I take that back - the bounce stats data was 'lost' from my display for a couple days, but the retrieved it and added it back in. Happened once, IIRC - generally nice people, and there's an IRC channel for help if you want that (in addition to standard other help channels).

Tried critsend.com -- the service threw up random errors and didn't really work well for us. In the end, went with SendGrid. Though if you are sending <500 emails a day, using Gmail SMTP servers is the best option.

Postmark App (http://postmarkapp.com) has worked well for me in terms of sending large quantities of emails in the past. Easy integration, good team, active development.

I've just used PostMarkApp on a site, I love their interface, all works great but they only allow 'transactional' emails, which I guess is like 'click here to activate your account', whereas my site was sending out alerts to people (kind of like facebook does, but not as often), which they didn't seem to like too much.

Does sendgrid allow this? I've signed up to SocketLabs which seems to do it but haven't been able to get SMTP sending working correctly yet, haven't spent much time on it though.

For the record, the amount of hassles I've had with SMTP servers these services are very handy, especially for logs etc like postmark has

In our usage, Sendgrid never complained. No particular knowledge about specific usages.

wait, what do you mean Postmark doesn't allow sending "alert" (notification) emails? Can you elaborate?

I've just started integrating postmark on an app I'm working on; their API & docs are excellent, they're not too pricy either.

I used to use authsmtp.com before switching to SendGrid. They're okay, but SendGrid gave me better deliverability numbers and a lot more features.

The two I love most on this list are SendGrid and Dropbox. Both solve a real pain point in a simple way.

Blog is down... "Site Error: Unable to Load Site Preferences; No Preferences Found"

Should be OK now. 600 concurrent users is big for us.

I like getsatisfaction. It makes it really easy to aggregate user voices about your product. I also use Mailchimp for a couple of clients - highly recommended.

I recommend Piwik for real-time statistics alternative to Google Analytics. It's free also.

Yeah, office Glico is the besssssttt!

Finally, some love for the Gleec.

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