I’ve been experimenting with Bookbub ads over the last few weeks, basically gambling small sums to advertise my novels on this huge email-based book marketing platform.
Like many online ad platforms, Bookbub is based on an auction system: Bookbub puts your ad in front of eyeballs, you make a bid for the owners of the eyeballs to click on your ad, and you hope that the clicks turn into purchases.
Your success depends on a lot of variables interacting to hook the right readers. For example, you can target readers who like particular authors, who live in different countries, and buy from different vendors (e.g. Apple, Amazon, Kobo). You can vary your price-per-click bid rate, and the price of your book. You can advertise on different days of the week, and you can choose to release to spend your money quickly or slowly. And of course, you have to design an ad that will seduce those eyeballs. The ad on the left has been one of my most successful.
There are lots of whizz-bang guides on how to get the most out of online ads (even entire courses!), but the take-away message is that you must test multiple versions of ads to find the optimum combination of variables.
As a former academic who has crunched a lot of slippery data (my field is linguistics), this kind of testing looks very complicated; there are so many variables. It is made more difficult for a small player like me investing an average $10 a test because the scale of results is too small to be statistically reliable. For example, why was the clickthrough rate for Apple and Kobo higher this Friday than Thursday, but lower for Amazon? (See the graph at the top of this post.) If I invested $10,000 rather than $10, I’d be much more confident in the results.
So where to? Short of running hundreds of small A/B tests or performing a factor analysis on a $10,000 test, I’m falling back on the approach I used as a linguistics academic when I operated in the grey zone between qualitative and quantitative data: (a) Start with a rough working hypothesis (b) gradually modify the hypothesis as new data comes in, (c) test the modified hypothesis. In practice this has entailed about ten tests so far.
And what have I Iearned? Well, here are some trends, but bear in mind that context is everything: I’m a ‘mature’ male Australian hybrid author writing quirky espionage fiction, and psychological and satirical thrillers, not a young female American author of time travel shape-shifter romance.
AD DESIGN: A quote from a review works better than a summary of the plot.
REGION: Australia (and to an extent Canada) are less competitive than the US and UK.
VENDOR: Apple and Kobo sell as well as Amazon in Australia.
WEEKDAY: Weekend ads may do less well than weekday ads.
AUTHOR: Readers of Daniel Silva like my books.
Remember these are trends based on small stats, not firm conclusions.
The outlook? I’m getting close to a return on my investment on my ads. I’m planning to run my best ad with a bigger investment, but to test some variations with a simultaneous low-cost ad.
I’d be fascinated to hear from other authors who are treading the same path.