Hacker News Reader

Comment by patio11 | Hacker News Reader

I really cannot emphasize enough how the intersection of these two fields is a) extraordinarily rare, b) extraordinarily capable of producing directly attributable, measurable improvements across an entire business and as a direct consequence c) very, very richly valued by the market right now.

View on HN

>c) very, very richly valued by the market right now.

Yeah. I think, in fact, it's overvalued by the market right now. Many people are bidding for ad words as if the google adword is what brought you the customer.

That's the thing; because the adwords are the first hits and are super convenient for the customer, they pick up a lot of people that would have clicked on the organic result.

Now, this is okay if they are taking a very small cut, but the way many people use google adwords? the price is bid up to a very large cut of the profit you are going to earn from that customer. I mean, yeah, sure, if google really is bringing me brand new customers that had never heard of me; people I would not otherwise have a chance with, yeah, I can see paying 90% of the profit. But if google is bringing me customers that were very likely to come to me anyhow? eh, I'm not going to want to pay google nearly that much.

I think this is also why facebook is so highly valued; people believe that facebook has enough tracking data to see that blog that my customer read last week that actually caused the customer to sign up through the adword this week. And maybe they do... but it's a dang difficult problem, and I haven't seen facebook make much progress. Even then, they are going to have a hard time integrating the people that heard of me because of the checkout-line advertisements I bought in Mountain View Safeways.

(because of this, my advertising dollars go into cost per impression advertising, rather than cost per click.)

While I do not disagree with what you have to say with regards to Google getting to claim credit for skimming the cream off of their advertisers' own brand equity (and then selling that cream to them at impressive prices), that's sort of orthogonal to the point I was making. Specific to, say, AdWords, the point I'm making is that people who a) have enough development capability to ship software products and b) also know AdWords well enough to get entrusted with a company's budget are as rare as hen's teeth. Because many, many companies have 6 figure AdWords spends monthly and generate millions of dollars of revenue as a result of them, someone who is capable of applying their skills horizontally across a company's AdWords operation (+) can be very, very useful to have around.

There's approximately four people in the world that can do that. Two of them like their jobs, one just got funded, and the fourth one can charge whatever he darn pleases. (This is an exaggeration, but not much of one.)

+ You might naively assume I'm talking about the AdWords API, but there are so many things a developer could do to wring extra money out of an AdWords campaign that it is virtually cheating. A consulting client once bragged to me that they had a lot of landing pages, and they asked me how many I had. It turns out that they had 11 and I had, at the time, only 900-something. There are a lot of marketing departments which think that developing 900 landing pages would take a team of 20 people the better part of a year. A properly motivated intermediate Rails developer could bang that out on any given Monday. (P.S. If you're not going to make the company a million bucks next Monday, consider banging this out.)

>Specific to, say, AdWords, the point I'm making is that people who a) have enough development capability to ship software products and b) also know AdWords well enough to get entrusted with a company's budget are as rare as hen's teeth. Because many, many companies have 6 figure AdWords spends monthly and generate millions of dollars of revenue as a result of them, someone who is capable of applying their skills horizontally across a company's AdWords operation (+) can be very, very useful to have around.

This is something I have very little knowledge of... I have a (perhaps irrational and unjustified) generally low opinion of marketing people, and little contact with that world, but I thought that adwords/seo people were usually mediocre programmers.

> (P.S. If you're not going to make the company a million bucks next Monday, consider banging this out.)

so, how come you aren't a millionaire many times over? Clearly you are good at this sort of thing, you are a reasonable programmer, and on top of that, you are a good writer, and have a bit of fame. (I am convinced that being able to write well is one of those skills that, while nearly useless on it's own, dramatically multiplies the value of any other skills you obtain.)

I mean, I'm not saying you are wrong... just that it's more complicated than that. Marketing fluff, I mean, selling something for money that is pure marketing and nothing else is harder than it looks. And actually providing a service to a large number of users is also harder than it looks.

There is always that choice of "do I spend my time improving the product?" or "do I spend my time on marketing?" - the more time you spend on the latter, the harder the former becomes. The more time you spend on the former, the easier the latter becomes.

This is my current problem; I can't add users much faster than I am without degrading the service for existing customers. Marketing, right now, is mostly a 'for fun' thing.

You're mostly right. SWAG numbers: 50% of SEM/SEO professionals are FizzBuzz-level efficient in Excel, 10% of them know what a pivot table is, 5% of them know enough scripting to do their jobs. Against that expertise curve, the least accomplish programmer in e.g. an entire class of YC startups is a programming god among men. If you can ship products, you're so far ahead of the curve it isn't even funny.

so, how come you aren't a millionaire many times over?

The things I'm really good at tend to scale with the size of whatever line-of-business I'm allowed to work on. A 10% improvement in BCC's numbers is worth, ballpark, $5k a year or so. (Or it was before I doubled conversions recently -- guess it would be worth $10k a year now? Ask me in December for a better guess, summer throws things off.) A 10% improvement in $CONSULTING_CLIENT's $PRODUCT is worth, ballpark, a lot more. The large majority of that upside will accrue to whomever beneficially owns $CONSULTING_CLIENT.

marketing fluff

You don't need to have a product of your own to sell to monetize an engineering/marketing skillset if you (or someone in the audience) hypothetically has it. Many companies already have things to sell, and they will pay handsomely, in a variety of ways, for your expertise in making that happen.

This is my current problem:

I'd tell a client in your circumstance to focus on the hardware provisioning or CS bottleneck or whatever which is apparently making it infeasible to service VPS customers acquired at scale. In the meanwhile, continue experimenting with marketing channels. After solving the bottleneck, double down on whatever marketing channels seemed to work in microscale.

Yeah. I think, in fact, it's overvalued by the market right now. .. .. This is something I have very little knowledge of... I have a (perhaps irrational and unjustified) generally low opinion of marketing people, and little contact with that world, but I thought that adwords/seo people were usually mediocre programmers.

Um.

It might be worth holding off evaluation of how much something is worth until you know what you are talking about.

Two questions, why do you say there are only four people?

Also what is your advice about the landing pages?

"Four people" is an exaggeration intended to be humorous. The more straightforward way of saying it is: there is tremendous market demand right now for this skillset. This skillset is extraordinarily uncommon irrespective of the demand for it. Additionally, market conditions currently offer many independent, extraordinarily attractive options to individuals who possess this skillset. Three options, among many, include being a) extraordinarily highly compensated at a Fortune 500 firm which is capable of being arbitrarily generous to retain their services, b) using their skillset to found a startup, or c) forsaking full-time employment and instead consulting for numbers which would make your head spin. There exist other options, too. This means that the number of people in the (already small) hiring pool who are actually available for hire is even smaller than you would otherwise assume.

My advice about the landing pages is in more depth on my blog in multiple places, but basically, "Build a CMS to spit arbitrarily high numbers of landing pages out of some data source which you have and some templates. Iterate on the templates in a data-driven fashion." It's essentially as technically complex as the Rails 15 minute build-a-blog demo. For more elaboration, see blog.

what's your opinion on building automated computer-generated content from data you own (trends from stats, etc). Isn't that frowned upon by Google?

That's an oversimplification of Google's approach to mass creation of pages vis-a-vis organic SEO. Google will let (many) pages that would never appear in the main results appear in AdWords ads as long as one's credit card doesn't get declined. (There are things Google disapproves of in landing pages, but "having lots of landing pages" is not one of them.)

Do all these landing pages live on the primary domain or are they spread around different domains?

OK, so what about auto-generating pages that you intend to rank for organic search, NOT to advertise in Adwords? Do you think this should never be done? Or are there scenarios where it's gray-hat, but you figure it's safe to do?

You can always throw in some noindex,nofollow tags on the landing page to prevent Google from indexing them.

You could also consider using canonical tags for landing pages that are slight variations of each other [to prevent duplicate content penalties]. This one is really just a guess.

I thought you had some lady who writes your landing pages for you. Have you got it totally automated now, or am I confused?

Content for the website was written by freelancers, yes. It all sits in the database. My AdWords landing pages repurpose a subset of the same data with different templates relative to the (publicly exposed) content.

Four people is just an exaggeration. He's saying there are very few people who are genuinely great at this stuff.

Sincere question -- how do you programatically create 900 pages that aren't crap? I often advise the multiple, targeted landing page strategy, but I've always considered the best performance to be carefully considered pages with crisp copy that should at least be reasonably performant, and then test the hell out of those pages to try and boost them.

I've seen the programatically generated 'query-catcher' pages that the big boys use and have always dismissed them as effective (because they don't work on me). Am I completely missing the boat somewhere or is there some intersection I've yet to discover?

There are many people in the world who a) don't even know what a landing page is, b) hypothetically, if they knew they were at a landing page, would not be able to tell if it was programatically generated and c) if they knew it was programatically generated, still would not care. All they knew is the Googles had the exact answer to their question. Thanks Googles!

More broadly, "My behavior is not a good proxy for the behavior of all people everywhere." and "I may be smarter than substantially everyone in my industry but that is not the way I should bet." are two lessons to learn early and well.

Heh. To the contrary, I was betting that if my simple-old self saw what was up, then so would everyone.

Thanks for the head's up though. I'm still all for the careful crafting of landing pages, but at least I know I wouldn't be wasting my time to script out a bunch to at least put up until / unless they can be replaced or upgraded.

I would advise a mix of both strategies. Auto-generated content can allow you to test a large number of keywords/niches, and you can go to town creating better versions by hand once something has proven itself to actually be worth the effort.

I don't suppose we could persuade you to link one or two of those programatically generated LPs, could we? I'd be really interested to see how you're doing what you're doing.

The problem you're describing is well known in digital marketing. It's called "last touch attribution" - basically giving the last touch point your customer had with your company all the credit for the sale when there often are multiple touch points (story in techcrunch, blog post on site, mention on twitter, then finally a click on a banner ad when the customer is ready to buy). Here's a blog post on hacking Google Analytics to help: http://www.seomoz.org/blog/how-to-get-past-last-touch-attrib....

A google search for "last touch attribution" will give you other resources. Agreed on it being a very difficult (and valuable) problem to solve, though. The privacy issues with tracking a person's every touch point with your brand makes it even harder.

Huh. cool. Of course, even if that problem is solved, it's still very difficult to track meatspace interactions (I spend, actually, more money on meatspace ads than on online ads. check out the checkout lanes in the safeway on shoreline in mountain view.)

If I'm right that people are willing to pay more for advertising they can track, and that meatspace advertising will remain nearly impossible to track, meatspace advertising will remain undervalued.

The thing is, at least in my case, this simply isn't true. Nobody buys a VPS on the first click; if someone clicks through a google ad for my company, yeah, google should get some credit, but most of the credit should go to the blog the customer saw a month ago. That's the thing; because the adwords are the first hits and are super convenient for the customer, they pick up a lot of people that would have clicked on the organic result.

This is called attribution measurement. It is generally considered the next big thing in advertising technology - primarily by anyone who runs a website where buying intent isn't obvious.

Facebook isn't valued because of this, it is valued because of the sheer volume of people. Facebook lets you advertise to people who "like" something, but doesn't do anything very smart to predict this (perhaps they might in the future).

Google, OTOH, does have data on what you search for and actively uses it on 3rd party sites, as well as offering re-targeting.

As a programmer/engineer/developer/etc. and marketing enthusiast, I agree with your post. I know I'm quite rare in regards to my understanding of both fields. I find more similarities that differences. Where some hackers might see marketing as voodoo, I see it in the same way I see code: Just a collection of steps that are meant to achieve something. One of the best examples I can share is the one of designing a marketing campaign using code. Yes, code. I sit down and just create a list of "objects" which are the marketing tools. Then I create the main program and write in the possibles scenarios using conditionals and loops. If I feel like having fun, I use Ruby's blocks (DO...END). It sounds really weird, but think about it. Marketing is just an object oriented program.

For example. One could sit down and write the following for an email mailing.

class Mailing:

    def __init__(self):
        self.name = "name"
        self.email = "email address"
        self.subject = "mailing subject"
        self.campaign = "mailing campaign name"

Then:

class WeekendSaleMailing(Mailing):

    def set_name(name):
    ....

    #basically define the mailing using code.

The you go on to __main__ :

    weekend_sale = WeekendSaleMailing()

    #set the details of the mailing (customer names, offer, etc.)

Then you go on and describe every scenario that you can think of and describe it using conditionals:

    if client responds_to_mailing():
        #...

    elsif client open_emails_but_does_not_buy:
        #you might want to then write more code describing what to do in this case.

You might also want to include a loop to describe what iterative process you will use to market to a customer.

For example,

    while customer does_not_buy():
        #send a different mailing or marketing tool.

and so on.

It may seem like a lot of work, and it is. But it just shows that marketing is just like code. In the way that you abstract away your marketing plan and design it as if it was a complex object oriented program. This way you still feel that you are programming and not outside of the scope of your skills.

Try it out. It works. And it takes away the sense of weirdness of marketing. After all, it is not marketing anymore, its code.

To add,

Same with doing marketing tools. I write my ads by using code. I have a basic template of an ad, and then I run through it the different headlines and paragraphs until I find an ad that works. Then its just testing to see which ads pull better than others. Testing is super easy because you now have a framework to work with. You dont go blindly into it, you have a set of parameters to measure. For example, Headline 1.a might pull more than headline 1.b, because it added the more "FREE" at the end, instead of using it at the beginning.

Anyhow, what patio11 says is true. If you approach marketing as if it was software, you are going to really get ahead because most people treat marketing as if it was black magic. It isnt.

email in profile for questions

PS. Its not real Python, btw. Just pseudo code.

Reply on news.ycombinator.com