Building a feature to figure out how much our customers want to chat with agents | Hacker News Reader

Building a feature to figure out how much our customers want to chat with agents

medium.com

28 pointsjgranof15 comments

The reason I prefer humans to AI or whatever is that humans are infinitely better at fuzzy searches. I'm not enough of an expert to know what I need to know to get what I want. So I relay on a human with experience to translate for me, then give an answer. This is also an issue I have with indian call centers. Its very hard to communicate something to them in a manner that doesn't use the "correct" words. So I lose a lot of the service I'm counting on.

To me there's no value in talking to a human if the human is roboticly following a script.

It's really frustrating to call the helpdesk and get:

HD: Hello! What's your customer number?

Me: xxxxxxx

HD: Thank you, what's the problem?

Me: It is failing to connect to the network. I've tried a soft reset, then a factory reset, and finally reinstalling the firmware. While sniffing the traffic it looks like the device is failing to accept the DHCP responses from my server, as it continually broadcasts DHCP solicitations and is send responses but never configures its network settings.

HD: Have you tried turning it off and back on?

I’d be interested in the stats on call resolution at the off/on step. You have to assume it’s large enough that they ask it first.

Also I’m genuinely curious, what sort of response you expected to get when you called and told them about traffic sniffing.

Speaking as a one-time 'high speed internet tech support customer relations specialist' working at a call-center contracted out to Comcast, I was told that it was close to 90%. That seems exaggerated since I would say it was closer to 70% (ballpark) success for my calls, but it was high enough that the place eventually implement an entire first-line crew that did no more than ask about whether power cycling had been complete or explain power-cycling before handing the 'sub(scriber)' off.

If someone called in with the GP's depth of troubleshooting completed, the standard policy was say something like 'this is outside of my allowable offered support'(but if you really know how, help, but not officially!).

I'm usually hoping that they'll send me on to level 2 support so I don't have to redo all of the stuff I've already tried. I won't even try calling support until I've exhausted their FAQs/Knowledgebase and Stackexchange/Google. Picking up the phone is the last resort. Unfortunately it's fairly rare that the phone support is of any help.

Giving the customer the choice of how to contact you will always be a winner. Some will talk, some will text, for some it depends based on what's going on. But giving the customer a choice -- and respecting that choice! -- is always good for the customer.

This makes me giggle because there's a very simple heuristic that gets you a 99.9% solution at least:

Assume the customer does not want to talk to bots. That includes bots implemented on a human substrate, i.e., call-center scripts.

This is true. We've got a human in the loop, who is primed with alot of context around your request. Customer service + advice on coverage is the primary thrust of the feature.

Greetings and salutations! Welcome to the emergency line of the San Angeles Police Department. If you'd prefer an automated response, press 1 now.

Unrelated, but I f*cking hate Cover. I tried getting a quote with them once and they gave me some quotes that were not what I wanted and had budgeted for that policy. Not a big deal.

A week later I started getting bombarded with junk mail sent to my physical address, from insurance companies offering me every possible insurance policy in existence.

What a good way to innovate./s They claim they're modernizing the process of buying insurance but still use the same old trashy tactics of the old companies like selling your personal information.

CEO of Cover here. This is super surprising, as we sell our own policies + own a national insurance brokerage. No need to share/sell your information at all (nor would we, as a matter of principle), because we have access to most carriers in the US. We'd prefer to sell you a Cover policy or one from our partner panel that can be managed via our service.

Do you mind sharing which companies sent you physical mail? I'm at karn at cover.com. Would be great to dig in.

I don't remember exactly since this was a year o more ago. A lot of companies! I only remember talking with the Cover agent via text message.

I was quoting for basic Renter's Insurance. She gave me a price of 125/year which was too much since I only need it super basic insurance and both Lemonade and Assurant give you that for half. Anyway. Not a big deal. Exactly one week after that I started getting junk mail from insurance companies.

I never quoted with anyone else besides Cover. I bought from Assurant after quoting with Cover. Cover was the only possible culprit for that burst of junk mail.

I don't know how you guys share data with other companies or vendors but definitely something happened in the way my info was handled. The timing between that interaction and the correspondence I started getting made it pretty evident.

Are you sure it was Cover and not another source? Did you search in your browser, Google, etc? Was your existing policy expiring around the same time? Basically whenever you go shopping and don’t buy, an insurance carrier or agent can assume you bought elsewhere. Be prepared to get communicated to in 11 months. Generally the party you went to directly (agent like Cover or if you went direct to a carrier) owns the relationship so a carrier won’t market you unless you went directly. It’s pretty common for marketers to be able to target you those and many other ways.

Also, most (if not all) insurance carriers do not sell personal data and in fact go to great lengths to protect your data. It’s against the law in some cases or breach of contract between agents like Cover (which is probably why Karn is asking who you received mail from).

Disclosure: I work at an insurance carrier (not Cover)

I mean. It could have been a very unlucky coincidence but that would be unlikely.

I was trying to buy new renter's insurance, since the previous one got expired like 3 months before that. So I was for 3 months without renter's insurance, when I quoted with Cover. That's just to say that there was no relationship between the date of the junk mail and the expiration of the previous policy.

I got the quote from Cover, because I was seeing the ads and thought they could offer me a good deal. A week later I started getting this insurance junk mail.

Yeah if you were seeing ads for Cover then you’re already likely in someone’s “high propensity to buy insurance” list that ad bidding platforms use to target. Insurance companies usually deploy digital marketing capital fairly efficiently and hunt for people who are in the market for insurance. So once you clicked Cover’s ad then likely the ad platform tagged you and you ended up in target lists.

There’s a few other ways, like soft pulls against your credit report that usually flag you as interested in insurance. Some companies make it their business to know when your insurance expires. There’s lots of data brokers out there mixing data in a pretty invasive way that will get you into lists.

Reply on news.ycombinator.com