Optimizing Straight For Checkout
Facebook is always adding features and I write about them from time to time. These are headlining features that they write about in their blogs.
But sometimes we need to be reminded that at the same time they are doing this, they are bringing out new improvements behind the scenes. You can only notice these though through day to day experiences. One of the biggest behind-the-scenes improvements I noticed over time is the improvement Facebook has made in their website conversion targeting systems.
Before
In Ecommerce, you always have a funnel. It starts out with a visit to a landing page, then you move to some small commitment like an email signup or an add to cart, and then move into checkout and purchase.
The conversion rates for each of these events vary.
- Visit your website (n/a)
- Your customer commitment (10 to 15% conversion rate)
- Move into a checkout stage (5%)
- Completed checkout and purchase (1–2%)
Because a purchase and even a checkout is so rare, my operating principle had been that we cannot go “straight to the source”, optimizing the conversion algos for a Purchase. How can we expect Facebook to sift through its millions of people to find that special 1–2%?
So instead of that, we told the algos to optimize for an email or an add to cart, which have a higher conversion percentage. Then we would pray that once people have added something to the cart that the website is capable of moving them to the purchase event. This was a workaround but I felt it to be a necessary one.
After
I was talking to a few advertisers who I respect greatly. Whenever I get that chance, I love to ask them about what is working for them right now. Nothing surprising until I asked them about their optimization settings.
They were optimizing directly for checkout. I found this odd and attempted to correct them. Nope. No more Add to Cart nonsense. Not even Start Checkout. They go right for the Purchase.
So I went to one of the campaigns that has been a trouble for me a long time and switched the settings. I found that it worked. After 2–3 days, I found the cost of purchases — the metric we want at the end — had started to go down. It felt like a miracle. This truly would not be possible a few years ago.
What to do
I am not sure how deep these improvements are. This particular campaign has had the benefit of “sampling” a lot of purchases made organically on the website. I think if you are starting from scratch these improvements are not going to give you much benefit. I am thinking maybe 100–200 purchases is enough “indexing” when the targeting models are good enough that you use them.
But the fact we can even discuss using it in this way is pretty cool. No more spraying Add to Carts and praying that they go through. Optimizing straight for Purchase is a viable option.
Originally published at AdsJunket.
