Stu Smith: Making It Up As I Go Along
“ My life working for BinaryComponents, coding, design, and other stuff. ”
Running advertising is all very well, but unless you can track its effectiveness you could be throwing money down a drain and not even realizing.
We're a Microsoft coding shop here (not for any particular reason other than that's where our expertise lies), so we need to track IIS activity. It's easy to set it up to dump out plenty of information, but tracing through all that data needs automation.
... Enter LogParser
For those who haven't heard of it, LogParser is a terrific little free utility that can analyze logs of all kinds, not just IIS ones. It's a Microsoft utility, but it seems to have been disowned by them, however a thriving community has sprung up around it.
The beauty of LogParser is that it presents a SQL-like interface to flat log files. You can query, sort, and group the data in your logs, and then output in a variety of formats, including tabular and graphical.
A little example is in order.
One set of data we want to know is where people visit our site from (referers). This can be accomplished in LogParser using the following SQL:
SELECT TOP 20
EXTRACT_TOKEN(cs(Referer), 2, '/') AS Referer,
COUNT(*) AS Referrals
INTO TopReferers.jpg
FROM Logs\*
WHERE date >= SUB(SYSTEM_TIMESTAMP(), TIMESTAMP('8', 'd'))
AND Referer IS NOT NULL AND Referer <> 'www.feedghost.com'
GROUP BY Referer
ORDER BY COUNT(*) DESC
Or in English;
"Take the top 20 referers (taken as the domain name, i.e. after the second '/'), place into a graph image, with raw data taken from the logs, in the last seven days, ignoring navigation within the site, ordered by descending count".
Which gives us:
The top referer (on the left) is Google ads, and the next two are ordinary Google searches and a review from Eric Nelson (which just goes to show the power of blog reviews).
Our log analysis now runs nightly and produces graphs for various statistics, together with detailed pages listing more items, which are then emailed to us with thumbnails. We currently analyze referers, downloads, downloads by advertiser, slowest, and errors, and we have more planned.
The next important piece of tracking, which we are working on, is tracking how many people we lose at each stage from advertiser to payment.
We lose 98.5% of people in the transition from the first to the second stage (which is to be expected -- how often do you click adverts?) We're experimenting with different ad variations here. At first we didn't make it clear that our product isn't free -- while this might have given us slightly more clicks, it leads to reduced rates downstream. We now include the price and the trial period in the advert, so we'll have to see how that works.
Stage 2 to stage 3 loses 80% of customers. Hopefully the advert changes will improve that, but we're also going to completely revamp our front page, as it's really pretty uninspiring at the moment.
Checking install and first run rates is very tricky as we've now left the website. We now launch a web-page on uninstall, asking for comments, in the hope of getting some feedback there. (And in fact, we've had feedback about a very nasty bug where FeedGhost didn't run under certain locales, so that was definitely worth it).
Stages 6 and 7, registration and payment, are obviously easy to track -- but they have a 7-day and 30-day delay in the normal case, making the figures difficult to track in the short-term.
Obviously we're in the early stages of getting our promotion working effectively. I think the key message is to keep watching what's happening, and change what isn't working.
It's been a hectic couple of weeks here at BinaryComponents as we've finished off our most recent big feature (tagging, link-blogs, and communities) and are gearing up for our first advertising push. I thought I'd write a mini-series of articles describing our experiences of the move from beta software to full release.
We've started off by placing adverts with various providers. We have plenty of other ideas for promoting FeedGhost but paid adverts are relatively low-effort, in that assuming you have some nice punchy phrases, you can get started with nothing but a credit card.
Google is obviously the big-hitter here and advertising is a cinche to set up with them. Payment is per-click and you choose your daily budget: once you use that up, your ads simply stop running. Pay-per-click also means that if your ads are ineffective (for example if you haven't written them well, or it's a product no-one wants) then all you've wasted is time.
Google goes all-out to help you produce the best advert possible: after all, that's how they get paid. The forms for entering text are great with a real-time display of how your ad would look, and when choosing keywords, Google will suggest others that might work, plus it gives an idea of what you'll be getting for your money.
Once your ads are running you can watch the number of impressions, the number of clicks, the click-through-rate, the cost-per-click etc, in summary form or broken down by keyword or by ad variation (you can run different versions of your ads for free to test their effectiveness).
We're averaging click-through rates of up to 1.4% for our best keywords, and given that our ads can be show hundreds of thousands of times per day it's definitely worth it.
Summary: Two thumbs up.
Text Link Ads run on a flat price per 30 day ad run. I personally consider this more of a gamble: if you have a great ad, you might get loads of clicks resulting in a very low effect cost-per-click. On the other hand, if your ad doesn't work, then it's really money down the drain.
Ads are placed on particular websites that you choose yourself from the list available. Unfortunately TLA don't provides links to the websites, so you'll have to search for them. From sampling of a few of these websites I found that either I couldn't actually find any ads on them, or the ads were way down the page. As an example, I looked at the top three "Top Ranked Text Link Ads" and I couldn't see a single advert on those websites.
If this were a pay-per-click service we might have gone for it, but I'm not paying for ads if I can't even see them.
Summary: No deal.
FeedBurner places ads in blogs, and charges an amount per 1000 impressions (CPM). FeedBurner divides ads into two categories: networks or channels. We very nearly didn't use FeedBurner as we started looking at channels and were shocked to find the minimum budget was $12,000 for the computing and technology one. It's a real shame that there's no entry-level way of getting channelled ads here.
However we tried again, this time in 'networks'. Here you choose from out of about a dozen 'network' blogs, ie blogs written by more than one author. We've only dipped our toes into FeedBurner for the time being; we'll have to see how it actually performs. Our ads went live today so no useful stats as yet.
One really annoying bug in FeedBurner is that as you enter your advert, it displays what your ad would look like - except it doesn't handle punctuation correctly, so you have to modify your sentences so that you don't end up with a full-stop falling onto a new line.
Unfortunately I can't find any ways of tracking click counts or click-through-rates on FeedBurner, which means that we're left to our own devices to work out if it's going to be cost-effective to continue. It seems that FeedBurner have been concentrating on the blog publisher side of things, and have neglected the advertisers who actually pay. Incidentally I'd love to be proved wrong about this; it might be just that I can't use their website.
Summary: Maybe.
Pheedo provides pay-per-click ads in blogs, and like FeedBurner place their ads in an impressive array of blogs. You choose the category you want your blog to show in, plus your budget, and away you go. They have a similar dashboard to Google that displays your views, clicks, and click-through-ratio.
One downside to Pheedo is that it's very US-oriented: fine if you live there, but I resent having to make up a state and ZIP code when I don't have one, especially since I'm giving my credit card details.
We've only just started using Pheedo, so like FeedBurner, I'll keep you posted as to how it works out.
Summary: Maybe. Don't forget there are people outside the US!
I happened across Project Glidepath today via Eric Nelson's blog, and while I'm not interested in a free Zune (other than to sell on eBay), I would be interested in getting FeedGhost spotlighted there since we've put in a fair amount of effort making sure that FeedGhost not just installs and runs on Vista, but also takes advantage of the glass effects.
Unfortunately the Glidepath site is, well... not up to Microsoft's usual standards. It's fairly difficult to work out what they're actually offering, but in the end it seems that they provide an add-in to Visual Studio that delivers content about starting a MicroISV via RSS.
Just a few problems...
I do appreciate that this site is probably a small start-up itself, albeit within Microsoft, and I'm sure it's a very clever little extension, and probably the content is useful, but I think I'd sum it up like this:
Glidepath might well have some great content there, but if you want me to install three different packages, risk losing my work environment, and lock me into a proprietary content distribution system that exists only to deliver one feed, well then not even the offer of a free Zune and some advertising is going to make me find out.