5 ms·

I'm about to release a bot for Whatsapp for NanaGram (https://nanagram.co https://nanagram.co). Amped! The user experience of SMS-first products is novel and f

by aacook 7y ago

I'm about to release a bot for Whatsapp for NanaGram (https://nanagram.co https://nanagram.co). Amped! The user experience of SMS-first products is novel and fun. There's something magical about snapping a photo and just texting it to a unique NanaGram number powered by Twilio. It takes just a couple seconds. I'm excited to deliver the same experience through Whatsapp to customers around the globe and people on wifi. Building NanaGram SMS-first has been a challenge. There have been all kinds of gotchas: 1) AT&T started blocking any number permanently across its network last January if it contains a hyperlink or email address, even if the customer interacts with the number. I had to create separate messages just for the 12.5% AT&T users. 2) Twilio only works when you have cell service. If you're indoors with wifi only, the text won't go through. 3) Images are downsized in an unpredictable way by the carriers. If you're sending 1 or a couple images, they're pretty good quality. If you send ~5+ photos, they get downsized pretty far and luckily they're still good enough for 4x6 photo printing. Certain carriers like cricket are super aggressive with downsizing. 4) When Apple users with iCloud enabled share more than 1 image, it generates a sharing album which has been a beast to support. Then there's the added challenge of Apple's recent move to HEIC/HEIF images. 5) Twilio doesn't support group threads (WhatsApp doesn't yet, either) ... I could go on and on. Getting approval from Whatsapp took a while. I first applied about a year ago and didn't hear anything, then applied again and was approved a few weeks ago. I'm guessing the approval layer is to ensure only high-quality bots make the cut. They also have an approval layer for all of your messages sent outside a 24-hour window of the last user interaction. The approval for that took a few days. Integration with WhatsApp has been surprisingly easy. I basically just had to change "+1{number}" to "whatsapp+1{number}"

pricechild 7y ago
> 1) AT&T started blocking any number permanently across its network last January if it contains a hyperlink or email address, even if the customer interacts with the number. I had to create separate messages just for the 12.5% AT&T users. I don't follow - they blocked any number which replied with a hyperlink or email? Can users of AT&T not send them to each other?
- aacook 7y ago
  If you send a message containing a hyperlink or email address from a Twilio number to an AT&T number, AT&T will block that Twilio number across its network without any notification. This happens even if you have an ongoing thread going with the customer. Even if after the block happens the customer pings the bot with "START" or any message, AT&T silently ignores it and continues to block the number. The customer can continue to send messages inbound to the number (and I believe those get received by Twilio but can't quite remember) but any outbound responses are blocked by AT&T from making their way to the customer, even though they're marked delivered by Twilio. This was the case as of Feb 2019. It could've changed since then and I haven't checked recently.
- filoleg 7y ago
  I bet that AT&T can detect "non-real-user" phone numbers, so their hyperlink filter only acts on those numbers. I have been using Google Voice as my secondary phone number for the past 5 years, and some services manage to detect that it isn't a real phone number and refuse to work with it. More specifically, it usually happens when I try to use 2FA over SMS (note: i know that 2FA over SMS is not considered secure, but some of the services I use only offer this as their 2FA, and it is better than me not using 2FA at all). It doesn't just fail to deliver the message, the service straight up tells me to enter a "real" phone number.
  - aacook 7y ago
    I've noticed this as well. Services like Twilio provide an endpoint to check the carrier. For my Google Voice number, they return "Google (Grand Central) BWI - Bandwidth.com - SVR."
rob-olmos 7y ago
Did Twilio or AT&T ever provide any response on the blocking numbers issue?
- aacook 7y ago
  No, they don't. I haven't tested this in a while though. Edit: Oh, I think you may mean a support response. Twilio helped me dig into the issue to great lengths. Ultimately, it was out of their hands. I think it's a lazy decision on AT&T's choice. Seems like a very rudimentary spam filter. You'd think they'd at least whitelist numbers customers are actively texting, especially if they don't contain keywords like "stop" or "spam."

by aacook 7y ago

pricechild 7y ago

> 1) AT&T started blocking any number permanently across its network last January if it contains a hyperlink or email address, even if the customer interacts with the number. I had to create separate messages just for the 12.5% AT&T users. I don't follow - they blocked any number which replied with a hyperlink or email? Can users of AT&T not send them to each other?

aacook 7y ago
If you send a message containing a hyperlink or email address from a Twilio number to an AT&T number, AT&T will block that Twilio number across its network without any notification. This happens even if you have an ongoing thread going with the customer. Even if after the block happens the customer pings the bot with "START" or any message, AT&T silently ignores it and continues to block the number. The customer can continue to send messages inbound to the number (and I believe those get received by Twilio but can't quite remember) but any outbound responses are blocked by AT&T from making their way to the customer, even though they're marked delivered by Twilio. This was the case as of Feb 2019. It could've changed since then and I haven't checked recently.
filoleg 7y ago
I bet that AT&T can detect "non-real-user" phone numbers, so their hyperlink filter only acts on those numbers. I have been using Google Voice as my secondary phone number for the past 5 years, and some services manage to detect that it isn't a real phone number and refuse to work with it. More specifically, it usually happens when I try to use 2FA over SMS (note: i know that 2FA over SMS is not considered secure, but some of the services I use only offer this as their 2FA, and it is better than me not using 2FA at all). It doesn't just fail to deliver the message, the service straight up tells me to enter a "real" phone number.
- aacook 7y ago
  I've noticed this as well. Services like Twilio provide an endpoint to check the carrier. For my Google Voice number, they return "Google (Grand Central) BWI - Bandwidth.com - SVR."

rob-olmos 7y ago

Did Twilio or AT&T ever provide any response on the blocking numbers issue?

aacook 7y ago
No, they don't. I haven't tested this in a while though. Edit: Oh, I think you may mean a support response. Twilio helped me dig into the issue to great lengths. Ultimately, it was out of their hands. I think it's a lazy decision on AT&T's choice. Seems like a very rudimentary spam filter. You'd think they'd at least whitelist numbers customers are actively texting, especially if they don't contain keywords like "stop" or "spam."