There's something in "massively multiplayer online games" called theorycrafting. It's when you take a game and play it with an eye to figure out the numbers beneath it. You've probably seen Dungeons & Dragons played in a movie or television show, where dice are rolled to see if a player, say, escapes from a monster's grasp.

When you're playing with dice, you can see the numbers. "To escape, you need to roll greater than a fifteen on a twenty-sided die," your referee might tell you. When you play a game online, like World of Warcraft, the "dice" are hidden, the server is the referee, and some software is coming up with all of the percentages and values.

But "theorycrafters" want to see the dice. They'll play in ways where they can take known situations and outcomes, change a single variable, and then see how the known values change. They'll record things in spreadsheets, or even use programs written specifically to watch and record values as they play.

What theorycrafters are trying to do is, in effect, exactly like counting cards in blackjack. If you know what the system is doing... that there are, say, six decks in a shoe, and 13 kings have been played... your odds of winning, and what the odds are for winning what you want, go waaaaay up.

With online games, it's sort of like playing blackjack without knowing which cards were removed from the deck. The servers could be running whatever numbers they want.

Now, if you ask me, theorycrafting isn't fun. When I play, I want Blizzard (Warcraft's developers) to take my dice, please, and let me forget what I'm doing isn't much different from playing blackjack or slots.

Unfortunately, this attitude keeps me from participating fully in the game. In "raids", where 30-40 people cooperatively¹ fight the most powerful creatures in the game, everyone is supposed to be familiar with all sorts of theorcrafted knowledge...

• The best gear (armor, weapons, wraps, etc)
• Enhancements for that gear (armor kits, spells to make weapons stronger, etc)
• Character talents
• "Rotation" (which spells to cast in a fight and in what order)
• (No, believe me, we're just getting started, but that's enough for now... )

If you're not well-versed, and your character is wearing or doing the wrong things, especially in the toughest parts of the game, you're liable to get kicked out of the raid.² And those links I've included, above, were for just one "specialization" of the thirty-six in the game! Phew.

You get the picture. It's insane the amount of studying that needs to go on to simply know the results of theorycrafting, much less to do the theorycrafting yourself. And it kills some of the fun. Ultimately, there are other games I could be playing.

Which brings us to the lesson I'm trying to push here: Theorycrafters might be playing the same game as me, but being a theorycrafter is work, hard work.

Targeted advertisements are like theorycrafting. Discuss.

How does this connect with being an indie app developer? Well, MarkUpDown, my Markdown editor for Windows 10, is getting reasonably mature, and over the last year or so, I've had my ears open to other indie developers that talk about advertising. There's a decent amount of advice about how to market, but I noticed that I bristled when I heard about one kind of marketing in particular. It took me a while to figure out why: It's because there's a particular type of marketing that reminds me of theorycrafting, and I'm not convinced the juice is worth the squeeze.

I'm going to lump the advertising methods I've heard discussed into two categories: Broadcast and microcast.

(Dare you to figure out which one I like more.)

Broadcast advertisements

You're used to broadcast advertisements. Television, radio, and newspapers are the conventional examples. In a figurative since, there's always a lot of "bycatch" in broadcast. You're casting the net out broadly, knowing a good percentage of the people who are exposed to your ad aren't in your market.

You certainly want to cast the nets where you think you'll maximize the market you're after, but you're not being that careful. Fill it up!

Here's one example of a "new media" broadcast from John Saddington's blog, where he's talking about spending $9000 to advertise his Desk app on Daring Fireball:

So I began the paper-napkin math and figured that if I spent $9,000 on DF it would ultimately have to convert approximately 300 copies of my app to at least break-even. So I asked myself this simple question:

Do I really believe that Daring Fireball can create explicit value in converting 300 new customers?

News flash: It did.

The sponsorship went live on November 17th and I held my breath…

The short answer is this: Yes, DaringFireball worked for me.

Over the course of that first week’s campaign and sponsorship I had net (profit) proceeds of nearly $16,000, which means that my investment of the $9,000 was made back within the first week of sponsorship!

For those interested, the previous 7 days of sales were just north of $2,100, so assuming that I was on pace to do that again I was clearly in the black based on Gruber’s sponsorship.

But there were three more significant things that happened as a result of the sponsorship, the first is that I began to rank at the very top of the Mac App Store in not only the Productivity category but globally...

This means, at least in the dominant US-market, Desk App was the #11th most-grossing app in the entire Mac App Store marketplace. 

That's expensive, and risky, but fun.

An aside on advertising dollars

I can hear folks saying to themselves, "Nine grand?!! Are you OUT OF YOUR MIND?!!!"

One thing I feel I should point out: If you've written a detailed app, you've already invested lots more than nine grand in your endeavor. I don't want to think about it too much, but I probably have $30-40k of time in MarkUpDown. That's insane. It's a great Markdown editor -- the best on Windows, I don't mind believing -- and that's probably about what it'd cost to get someone with my experience to write it.

But that's a ton of cash -- or, to be kinder, a huge investment. If you won't drop $10k to get the word out on something that ran you $35k to write, well, your don't value your time enough.

I know, I know. As an indie, getting to write what you want is a huge part of the gig, and by "part", I mean "reward". And I know the feeling of wishing you could give some cash back to your employer to work on what you wanted to, even if it was constrainted to something in their app! I often feel that way about tech debt. Gosh, I'd work for half-price now to make my future work so much easier.

Regardless, the advice here is don't be penny wise with advertising and pound foolish with your time. $5-10k is not a lot to spend on effective advertising.

The question I keep asking myself is, "What advertising is going to be effective?" And with that, we're back to our consideration of...

"Microcast" advertisements

Facebook and Twitter, even Google AdWords to a degree, offer a different sort of advertisement than traditional broadcast. Targeted advertisements allow you to cast your net more precisely, reducing the bycatch significantly.

Now, instead of broadcasting to everyone at a website (can you imagine if Google search had static banner ads?), you get to wait until the user has performed some action that hugely improves the chances they're interested in what you're selling.

• Someone searches for "Windows Markdown editor" on Google.
• Someone on Twitter follows John Gruber and Paul Thurrott.
• Someone on Facebook is a software engineer and hosts projects on GitHub.
  • (You're not writing your READMEs with a Markdown editor? Ouch! There's an easier way! ;^D)

These are increasingly specific demographics. You're no longer broadcasting to a group that you hope largely overlaps with your market. You're lasering in on people you already know are your market, and hoping your net works.

Of course, this service is going to cost you more per impression too.

Theorycraft and Microcasts

The thing that bugs me is how theorycrafty app developers sound when they're working on their targeted microcasts. Like theorycrafting a game, finding the right target for your advertisements is hard work. I understand the appeal that if you get everything dialed in perfectly, you could theoretically take your hands off of that perfect combination, step away, ramp up your advertising budget, and hit Phase 3. Profit.

But it always sounds like there are too many dials.

Here's a quick introduction to how it feels to use Facebook ads from Charles Perry and Joe Ciplinski on the Release Notes podcast, episode Sorta Cool, Sorta Creepy, from back in December 2015 (link is cued up to 8:33 - transcriptions and any errors therein are mine!):

Charles: You know… I wasn’t really familiar with Facebook ads before I started looking into this, and looking for options. And in fact I didn’t even have a Facebook account. But they know an incredible amount about the people who are using Facebook. It is just mind-boggling. They give a lot of that information… I was going to say that they give a lot of the information to their advertisers, but they don’t. But they do allow advertisers to target based on that information.

You can target based off of geography or travel inclinations or whether they have a friend who just gave birth. It’s all kinds of crazy things that you can do -- you really can [target people with friends who just gave birth]. It’s really odd to have that much insight into… the people that you’re advertising to. It’s great, but it’s sorta unique in that you don’t get that much insight into who is viewing your ads in any other platform that I’m aware of, at least.

And from the same episode a few minutes later on...

Joe: So you can target pretty much anything you want, and I know it’s pretty crazy. We did a very brief experiment with Setlists a while back, which I want to repeat, because I don’t think we did it right. We were able to say just give us people who have iPads… Just give me musicians. We could even just say give me singers versus guitar players versus… It’s pretty wild how specific you can get in that detail, and that’s why Facebook’s always asking you all those questions like where you went to high school or where you live… but they also glean quite a bit from your timeline and from what people post about you and what you post in return.

Man, that's a lot of dials. And each one could be the barrier to hitting your target market. Who is more likely to use Bombing Brain's Setlists app? What if it's, strangely, drummers, even though there's a huge emphasis on chords in the app? You might never think to add them to a targeted ad, though a broadcast ad on The Garage Band podcast (made up name) would've hit them all. That is to say, microcasts seem to operate as opt-in operations.

Or here's Curtis Herbert's blog post, That Dirty M[arketing] Word:

One of the (many) things I tried was Facebook app install ads...

Over a one-month period I tried all kinds of permutations, re-visiting my ads every day or two to turn off the clear losers to save money, and try new ones. Looking back at the ad manager for that time period my worst permutation was going to cost me $20 a user, and I had a lot in the $2 - $5 range, but by Jan 2016 I knew how to dial some of the knobs get users for < $1.

That sounds like theorycrafting.

Or on the Under the Radar podcast, episode #49, App Store Search Ads:

Marco: Another thing to think about is what percentage of your new users come through clicking ads? If you're thinking that you make 25 cents per user, but 10% of you users come from clicking ads and the rest come from other ways, then you actually might be able to bid a lot more than what that is to get those users...

... I think [cost per acquisition] will make people think about their business models a little bit more and maybe charge more money and charge higher prices for apps that really, that can earn it, that deserve the higher prices. I think this actually might help raise App Store prices as people figure out, "You know, it'd be really a lot better if I could bid on these keywords and be a little more competitive on the ad side, but in order to do that I have to charge a real price for my app or I have to have some kind of recurring revenue scheme or something like that. Ultimately I think this might actually help ad pricing substantially.

Theorycraft or carpet bomb?

Look at the emphasis on cost per acquisition in the last two microcast quotes. That's the right metric, but I'd suggest "per user" is the wrong unit.

Maybe it's just mentality, but I think it's also largely medium. There are places your potential users live. If you can blanket one of those places that overlaps with your potential users closely enough, it seems worth it to me to broadcast there.

Now I figure part of the lure of microcast is microspending. You can spend $20 (or less) a month on these media if you want. If you target too broadly, your cash will be spent quickly, and you'll hit folks who weren't in your market. But if you start with the pico instead of the micro, you can work your way up, minimizing the cash you've wasted on bycatch.

But I also worry the time it takes to "pico-theorycraft"³ an advertisement... oh, the time. What features did you not add because of the time you spent theorycrafting advertisements? I think Curtis is happy with his results, but every time I have a process that complicated (including writing blog posts), I look up at the clock and say, "Where did today's time go?!"

And if the ads don't scale, even if our theorycrafted dials were working great, what time we've wasted! If I had to tune all those knobs over days or weeks, then cranked my spending from $20 to $2000, and didn't get 100x results, well, man. What a waste.

If you have a broad enough market, and the market congregates in a few specific places that are marketable, I think the lesson is that you go advertise on those media first. The broadest exposure is the best exposure, and has the best follow-on benefits. I also wonder how much of, say, Daring Fireball readers also would be the ones hit by a well-theorycrafted advertisement.

If I had a marketing department with people who really knew how to do this, I would expect them to make these experiments. When it's just me, well, easier is better. Time is money and all that.

But for a niche app like MarkUpDown, even though there was, a few years ago, at least one decent Markdown editor on Windows making enough money it got rewritten, I have to consider just how big my total potential market is. If I could reach everyone who would spend $23 on not just a Markdown editor, but my Markdown editor, how many would that be? If I could find Desk App's Daring Fireball equivalent, could I pull down 550 purchases? Or should I just consider this app a labor of love? If the market is thousands, and I can find the right source, broadcast makes sense. If I want to maximize the return on a labor of love, then I should keep up the hobbyist mentality and microcast.

(And then I think there's probably a middle-ground, where big companies carpet bomb with "targeted ads", but the ads really aren't all that targeted at all, since the markets are so large. "People who watch the NFL," doesn't really require theorycraft-level consideration, and if you're already advertising on TV during games, why not Twitter and Facebook too? Honestly, that's probably the best answer. If Saddington didn't market on Daring Fireball and then to Daring Fireball readers on Facebook and Twitter, he missed a great opportunity. If you buy into the whole saturation thing.)

In any event, there are my current thoughts on advertising, as I try to determine where to sink my hard-earned dollars so that they'll best make friends and bring more dollars home.

¹ No joke! From [wikia.com](http://wowwiki.wikia.com/wiki/Raid):

Raid groups are a way to have parties of more than 5 and up to 30 (40 for some old raids) people, divided into up to 8 groups of up to 5 players.

² Interestingly, as I've argued before, this means the game developers expect and effectively require you to play outside of the game to fully experience it. I'd argue that, additionally, Blizzard's level of financial success demands and depends on theorycrafters as well.

³ I realize I'm misusing "theorycraft" here. There are no hard numbers behind the scenes when you're talking about people. Oh sure, we can play amateur economist, or pretend our experiments in Facebook targeted ads are representative, or apply sampling statistics until we're blue, but we could be wrong too. See the major new networks from our last presidential election. That said, the methods between targeted ads and theorycrafters are similar, and the results could be too. Let's pretend the comparison works.

Labels: advertising, app dev, business, indie, markupdown