Hacker News Reader

Comment by blazar | Hacker News Reader

"An intern then manually clicked on each of [the 12000] addresses to ensure that the websites were ones the company wanted to advertise on." - true hero of this article

View on HN

Yeah, success of this entire pivot relied largely on this intern's audit report. I imagine that the intern had a massive list of 12,000 URLs and would click through them, taking care to note why the site passes muster or otherwise so that management can add a layer of justification to the final decision. Even at an average of 3 minutes per URL, that still amounts to >600 hours of repetitive and sometimes traumatising work. Remarkable that they only selected a single (unpaid?) intern to deal with this...truly remarkable.

3 minute per URL seem like a whole lot, it seems more likely that 7k of them were trash and were removed within seconds. A simple automated script/iframe with a "tinder" like interface of yes/no/maybe seems like the way to go here. Spend seconds on most, then minutes on the maybes.

You have a pretty optimistic opinion of the kinds of automation an intern gets.

Good point. Perhaps if you gave the job to a CS major, or an engineer they'd invest the time in making a time saving interface. But an intern in Chase? I doubt they would have the skills or the knowledge to do that.

Like, load the list into stumbleupon. I'm thinking about 6 seconds for trash, 30 seconds for interesting. So, maybe about 2 work days to remove trash and a few more to complete the rest. Could be done in less than a week.

You're assuming the intern actually did the job - that he didn't e.g. get bored a third of the way and finished the rest by dragging down the "No" field through the rest of the column in in Excel, and then changing a bunch of random URLs to "Yes".

My experience is that most people assigned with boring, repetitive jobs will either automate some of it or start cutting corners to avoid the work.

(In fact I'm starting to believe that the first runaway evil AI trying to take over the world will get stopped in its tracks, because it will not understand just how different the "on-paper" states of inventory and books in small and medium businesses are from reality.)

That's exactly what I would have done given such a task if there was no good way to automate it.

Many people do think exactly like that, but now realize that the output of such a task is often an input to another one. Garbage in, garbate out applies just as much to decision processes as to computers.

Most computer programs are just very quickly run decision processes, especially in the MIS subfield.

The task is "find out who is not providing value"

Your solution is to say they all provide value?

An as your boss I'd do simple random spot checks of some of those urls - you'd be under deeper review in a heartbeat.

The intern's requirements in this job are not "press every link presented" they are "find out who is not providing value for their worth" however the press dgaf about that.

The above is not advanced management techniques, it's basic check work.

Hey JPM has a meal allowance if you stay late. You can even stretch it to tomorrow's lunch if you're smart about what you get.

$25 - two sandwiches from Toasties baby :o

Casperjs to capture screenshot of every site, do a image match of the results for confirmation.

For better results, have that intern gather the email addresses of those 12,000 pre-approved sites, and email them a personal ad placement buy offer, or request a media kit.

You'll find all sorts of personalization opportunities and you get to cut out the middle-man.

It's kind of interesting that self-hosted direct advertising hasn't taken off as a technical problem to solve.

Because that's a business problem, the technical side of it was solved a long time ago.

I'm sorry, I don't understand. Are you saying that there exists direct advertising libraries for web development frameworks and CMSes, and that they've possibly been around for a long time?

Sorry for the slow reply. I'm not familiar with currently available frameworks/etc. so I can't speak to that side.

The point I intended to make is that, in my experience, there aren't site owners going "I wish I could do this if only it wasn't so technically hard", they're going "I wish I could do this if only it didn't mean people would spend less money with me". Because on the other side of the equation, ad buyers (generally, though obviously not universally) don't want to work that way, and they're the ones with the money and therefore the power.

I'm not a developer myself but I've worked with developers who created various in-house ad-tech, and while this sort of thing may be too complicated to do for Mr Random Joe Blogger without a CMS plugin, it's really not something that is too technically complicated to be feasible.

and now have 12000 billing portals to manage? 12000 different sites to update every time you update your ads?

Yes. This is how it was done during the magazine era.

Did anybody advertise in 12000 magazines?

If you include international magazines and newspapers, then yes.

Rarely. It was usually an ad agency that would do the actual deals to get things into magazines.

You could always contact a single magazine, but big corporations would let their ad agency handle distribution

Just give them your ad code (a JS snippet), this way you control your ads.

Billing is a solved problem too, no?

> Billing is a solved problem too, no?

No. Billing is a huge pain in the ass. Negotiating price with each site is a huge pain in the ass. Building a reliable JS snippet that works everywhere and that these sites will trust is also a huge pain in the ass. Everything about dealing with 5k sites manually is a pain in the ass, which is why ad networks exist in the first place.

actually a lot of them still use google, which can also run private deal ads.

Reply on news.ycombinator.com