Have you ever woken up to an ad account where everything is is on fire, and you want to kill everything and leave only the big ROAS numbers on the day of running?
One of the most common questions I get asked is that “Why is Facebook so volatile?” And this very question usually leads to one or more of the following thoughts:
- Is my ad account being punished by Facebook? Should I create a new one?
- I got one negative comment on my ad today, that must be throwing everything off.
- Should I 10X my bids because someone is obviously outbidding us?
Take a step back and CHILL for a second - take a sip of coffee, tea, or hit the vape, whatever you prefer. Mr. Zuckerberg is most likely not punishing you and the swing of performance is just... normal. Don’t start making abrupt decisions based on a limited dataset (ie. 12am - 8am on the day of).
To keep it CTC themed, here are some sports analogies:
1) If Lebron James struggles in the first quarter, but has averaged close to a triple double all season. Would you pull him out of the rest of the game and put him back in the next game? (Comment your favorite Lebron jokes below)
2) If Mike Trout bats 0/2, do you take him out of the game because he started poorly?
Chances are, most of you would say no to either one of these questions, unless there is an obvious reason on why they are starting the game poorly. The reality is that, we simply cannot and have no reason to make predictive decisions with such a small sample size, when the larger dataset paints a different picture.
EXAMPLE: ROLLING A 20 SIDED DICE
Here is a more dumbed down way to think about it. Below is a spreadsheet where I simulated the rolling of a 20-sided dice 1000 times a day, 7 days in a row. The 20 sides simulates an average conversion rate of 5%, and 1000 clicks a day. This mimics an account that is spending around $1000 a day if CPCs are at $1. If your budget is lower, expect higher volatility.
In this case, we chose the number 2 previously to represent purchase. The numbers in the table represents number of times the number landed during that day. The expected result is 50/day, but as we can see here that the actual results are fairly volatile. To paint a picture for the normal ad buyers mind:
Day 1: We are crushing!!
Day 2: A little slower today..
Day 3: Mark is punishing us
Day 4: Slight improvement, I’m going to dupe and pause the original adset to see if it does better.
Day 5-7: We’re back! Must be the duping and pausing I was doing.
The takeaway is this– performance fluctuations happen on a day-to-day level, we see it across ALL of our accounts. The bigger the spender, the less volatile it is due to sample size. Instead of freaking out and making irrational decisions with limited datasets (unless there are some obvious warning signs like your best ads getting disapproved), continue to focus on things that matter in which you have direct control over.