How do I know where to advertise?

This is part of an ongoing startup advice series where I answer (anonymized!) questions from readers, like a written version of Smart Bear Live. To get your question answered, email me at asmartbear -at- shortmail -dot- com.

Adventurous Advertiser writes:

I’m ready to spend my first $1000 on something other than AdWords.

There’s lots of options and I don’t know where to start. How do I pick?

Since you didn’t say what field you’re in, you’d think I have no answer, but I do!

There’s a thousand places to advertise and a hundred ways to do “lead-gen,” i.e. getting CSV dumps of contact information for so-called “pre-qualified” people who were tricked into downloading a whitepaper you’ve supplied.

Which are best? Are any actually good at generating sales?

At Smart Bear I tried literally a hundred campaigns: paid search, trade shows, postcard mailers, newsletter sponsorships, and magazine ads (yes, print!). I eventually built a system that could measure campaign efficacy with pinpoint accuracy, even with traditionally different media like print. (I once promised an article describing this system and have yet to deliver… I’ll remedy that…)

So here’s what I learned:

Nothing. I learned I have zero ability at predicting which campaigns would produce. The only way to tell was to ensure accurate measurement was in place and try it. And negotiate a lower “first time” rate for the trial.

Also the quality of the campaign would change over time. When I first started with Dr. Dobbs magazine I had a little 1/4-page ad in the back. (“In the zoo” I’d say to the ad salesmen, who disagreed with my characterization of their remnants.) It worked great, and a full-page ad worked better. But as the years wore on it became less and less effective, like clockwork, and I kept demanding a lower and lower price.

“But our readership is increasing” the ad salesman would say, pointing to their BPA stats (a 3rd-party “auditor” who always certifies strong growth and a monied readership). “Besides that, you’ll lose credibility if you stop advertising. People will wonder whether you’ve disappeared.”

The magazine went under a few months later.

But remember, three years before it died, it was one of the top lead sources. Things change. Can’t predict.

So here’s how you decide which to start with. Of the dozen possibilities in front of you pick the one that maximizes these things, in this order of precedence:

  1. Measurable — you have to know objectively if the value to the company is greater than the amount of money spent. Generally online is easier to measure than offline, but that’s not true for things like contests or special offers or secret landing pages.
  2. Quickest time to discover whether it worked. Fast learning means you get to a working advertising model faster.
  3. Cheapest. At this stage the valuable thing is seeking/learning what advertising works, not trying to maximize total revenue right out of the gate. Once you’ve got a handle on how to spend money to make money, it then makes sense to pour money on it as fast as possible.
  4. Comfortable — something you like, like a news site you already read, or a Twitter contest-runner that seems particularly smart about it. Since your gut isn’t too useful, you might as well pick something you like, especially if you already understand the culture of the site, and therefore could build an ad perfect for that audience. For example at Smart Bear I was particularly successful on StackOverflow, which I was an early beta user in, and on Coding Horror, which was was a long-time reader of.”

So don’t waste time fretting about picking the “right” campaign. You can’t. So make sure these key things are in place, and just go.

Add your advice to the discussion section!

17 responses to “How do I know where to advertise?”

  1. Oh, god, you had me at Dr. Dobbs.  I was cleaning out a bunch of junque last year and found my carefully saved Objective-C issue.  Swoon.

    I was still a subscriber when they went under and I think I switched my remnant to Computer Shopper.

    Yep, old.


  2. Great article.

    One other thought…

    Find the place to advertise that is closest to a relevant decision to buy. 

    When I was at Dell and ran university marketing in 2005, Facebook approached me to advertise. But why would anyone think of buying a computer when they’re trying to ‘hook up’ with someone in college class? The best advertising for Dell was CNET, where people were actually thinking about technology. The further you get away from intent, the less relevance and impact.

  3. I want to know how you measured print efficacy with pinpoint accuracy!

  4. This is actually a clear case of the Multiple Armed Bandit.

    Currently many people are calling the algorithm Epsilon Greedy (actually a broken version of it) MAB because they don’t know what they’re talking about.

    Multiple Armed Bandit is simple.  “Suppose you’re in a room full of slot machines, many of which are positive.  You don’t know the statistics.  What strategy will maximize your take?”

    Clearly, this is an adequate description of advertising.

    The epsilon greedy algorithm is one good approach.  (Weirdly, it’s being tossed around as an alternative to A/B testing, because again, the people who are doing this don’t know what they’re talking about.)

    Choose a threshhold, commonly 1 in 10.  For every lever pull, choose a random number.  If it’s under your threshhold, pull a random lever; otherwise, pull whatever lever has statistically had the best performance so far.

    Over time, all levers will be progressively pulled at some low rate, meaning even if you get bad pulls, eventually you’ll tend towards the good levers (though on big payout long odds, that can take a very long time.)

    Another good algorithm for approaching this is UCB-1, but it’s non trivial, so I won’t describe it here; the long and short of it is “use a confidence interval (the Hoeffding bound) to begin to trust your knowledge of certain levers, and drop pulling them off by the natural log of your confidence.”

    You’ll lose less in the long run, but it can be _really_ wrong on big payout long odds, so you need to be careful using it in some circumstances.

    The advantage of a multi-armed bandit approach is that it inherently keeps track of changing landscapes over time.  Dr Dobbs and Digg dead?  Reddit and Imgur doing better?  No worries: the algorithm will adjust on its own.

    The downside, of course, is that your marketing budget has to be large enough to encompass these fringe-case lever pulls in budgets their hosts will actually sell at.  Not hard on google’s $5 line; very difficult with campaigns that start at $1000.

    • That’s a good algorithm, and the one used by AdWords.

      It requires a few things to be effective, like immediate and dependable result-measurement, and an N which is sufficiently high for the number of tests being run.

      AdWords satisfies those conditions nicely; so do landing pages which are getting lots of traffic.

      Other things like “trade shows” and “magazine ads” usually don’t, so unfortunately you might need to use other measures.

      But thanks for the description — I agree this should be used far more often than it is.

  5. OK, this might seem obvious, but I try to never assume anything when sharing advice with start-ups. I think it’s important to start at the beginning, especially for those who might never have advertised before.

    1. Start with who you are trying to reach. It does no good to advertise in the “in” publication or on the “hot new” site if your target customers aren’t there.
    2. Then, determine where that target audience lives, works and plays. That’s where you need to be speaking to them.
    3. Once you know where they “hang out,” it’s important to give some thought to whether or not they will be receptive to your message, and most importantly, take action, when seeing your message in that location.

    For example, if I am a business owner you are trying to reach and I read a variety of business publications, as well as inflight magazines and leisure publications, which ones are most likely to be read by me at a time when I am receptive to what you’re selling? And, during the reading of which one am I most likely to act on your call-to-action? It’s not likely that I will act on a business decision while reading my leisure publications, but somewhat likely that I will act while reading an in-flight magazine, something I do (without being interrupted!)while traveling on business.

    The key is that advertisers must have a message that not only reasonates with readers, but that entices them to act.

    Of course, a clear call to action is essential for that to happen, and I can’t tell you how many ads I have reviewed that have no clear call to action.

    Businesses advertise for two reasons: 1) to entice action (make a sale, get someone to sign up for something, etc.); and, 2) to build a brand. If you have no clear call to action, your ad is simply an exercise in brand-building. And, if you have no call to action, how will you ever measure effectiveness?

  6. “Find the place to advertise that is closest to a relevant decision to buy… The further you get away from intent, the less relevance and impact.”

    I agree. Having a strong SEO background means I have “visitor intent” imprinted firmly on my retina. When you have limited funds or when you test new advertising channels you are seeking  measurable and immediate benefits (otherwise how would you know if they are promising?). If there is a mismatch between audience intent (not just audience type but intent) then you will not get adequate immediate returns (note emphasis on immediate).

    Having said that there are huge benefits in other types of marketing like inbound marketing but that’s not quick and may not pay off for a while so they are hard to test quickly. However reaching out to your target audience and forming relationships with influencers can only be beneficial in the long run as they not only have access paths to your target customers that are invaluable but may actually have valuable input on your product.

    Jason, I’m interested in how you measure the effectiveness of campaigns. In a world of multi-funnel marketing channels (multiple touch points before conversion) and delayed conversions, looking exclusively at conversion rate by source may miss the whole impact of the campaign (someone comes from campaign A, subscribes to newsletter, downloads PDF, then leaves only to come back refered by campaign B and convert – who gets the credit?). This is a common scenario which makes accurate attribution to campaigns difficult, increasingly so in a multi-device world.

    One theory I have is to look at the size of convertible pie referred. By “convertible pie” I mean referred  visitors who exhibit a certain behaviour making them likely to convert (now or later). 

    convertible pie = total visitors referred – visitors who bounce – visitors who have no intention to convert. 

    visitors who bounce = mismatch between their intent and the landing page/product.

    visitors who have no intention to convert = low engagement or have different purposes to accomplish on our site (we can assume that visitors are convertible if they take specific actions which we perceive as influential and leading up to that final conversion). 

    The reality is that a large proportion of traffic will NEVER convert. Looking at the size of the convertible pie within the entire traffic referred puts pressure on campaigns to deliver the best quality prospects (who match not only the traits of the target audience but the intent as well). It’s then the job of the landing page to maximise relevance relative to the campaigm and the rest of our website to do the selling and the converting. 


  7.  That $1k might be best spent on a Consultant helping you develop a marketing plan

  8. Hi, I’ve just read your post. Thanks for the tips that you’ve shared. I think you’re right. It is really important  to weigh things first before picking the campaign and your tips are helpful. :)

  9. try if you haven’t heard of yet. Btw great article :)

Sign up to receive 1-2 articles per month: