" Why do we bid on our brand?
Why do we show ads to people searching for our brand? Will people actually click on our competitor when they're looking for us? Time to find out.
Upsert Offices – Wednesday Afternoon
We're spending about $3,000 a week on brand terms in our Google account\nThe CEO is on an efficiency drive and wants us to turn this off\nI do wonder if these clicks are incremental: if someone searches for our brand won't they just click on the organic result anyway?\nJust to cover all the bases can you do some analysis to estimate the impact?
Should you bid on your brand terms? When someone searches for the name of your company they already know who you are. They had to learn your name somewhere, so it's highly unlikely that the Google ad they click on after searching your name deserves 100% of the credit. If you can turn off ads for brand terms you can usually save thousands of dollars to redirect into generic product terms or other demand-generating channels.\nThat said, there are some advantages to brand bidding. For example if a competitor is bidding on your brand term they might steal some of your best customers. So bidding on brand can defend your terms against competitors. In addition when running brand ads, you have more control over the message and destination. You can change the copy of your brand ad in an instant and A/B test different messages, which you can't do with organic search. Determining whether you should bid on brand terms or not requires a deprivation test, to find out how many of your sales are incremental, and how many would happen anyway.
Absence makes the heart grow fonder
4 weeks later
The brand campaigns have been off for a few weeks now, and you have the data to analyse. Choose two equal periods before and after the ads were turned off, and calculate the difference in total traffic.
When conducting a test before and after a big change, it's important to choose the same number of days before and after, to make it a fair test. Ideally those periods would cover the same days of the week, so your data isn't biased by the weekend effect. Also you should attempt to leave out the days in which the change was made, as new changes often cause volatile behavior: start a little after to give the change time to embed.
Before and After Test Results
All right. We've turned off our brand ads. How do we interpret the results? So you can see the ads are running for quite a while. And then around here we turned them off. We went from 56% share, 14% share, turned it off halfway through that day. So the share quickly dropped to zero as you know, all the kind of clicks from the campaign.\nSo to get registered in analytics and you know, there were no more. Euro codes trading around. So specifically paid traffic is from analytics. Anything that came through with a brand new campaign, a UTM a URL parameter, and then organic traffic is anything to the homepage from a organic SEO.\nAnd if you add them up together, that's total brand traffic. Cool. So how do we interpret this analysis? So let's look at let's, let's actually kind of rank this or show this visually in a chart so you can see what effect is happening. So I just selected it and then I went to insert chart.\nIf I get rid of total traffic we want to share this as a stacked area. And that's going to kind of add the two together. So you know, this red line here is the, some of the blue and the red. So, so you can see total traffic was like, you know, pretty spiky. There was a dip here, right when we turned things off.\nBut then there have been a few other spikes. So actually it doesn't look like, you know, the total traffic is down that much organic traffic actually. Pretty considerably compared to where it was before. So it looks like not all of the clicks from brand were incremental, which is exactly what we expected to see but how much?\nSo first we need to choose two different periods. We're going to work out the, to work out the traffic before the traffic after, and there's going to be averages. And then we're going to find the difference between. The other thing we want to be able to do is we'll get paid before. So how much paid traffic there was before?\nAnd then paid incremental. So how much of that was actually incremental? And then that, that should give us a good idea of what we want to you know, w what we want to decide about whether it's attendance back on. Okay, cool. So traffic before let's choose two equal periods, we're going to start on the Monday here at the beginning.\nWe're going to go Monday to Sunday, and then Monday to Sunday, again, we're going to take two weeks and that kind of balances out any you know, in any ethnic kind of fluctuations in the data Yeah, we don't have like say one weekend in this period and then no weekends in the period after because then if there's an effect on the weekend, you know, traffic's higher then it's going to bias the results of a test.\nWe also don't want to go straight from this Monday because you can see, we have at least one. Where the Brenda ads was still on and also you know, you can take, it takes some time for behavior to filter through the system. Just want to give it some breathing room and kind of you know, look at the data after the results of switching it off settled.\nSo we're going to skip this week and we're just going to go Monday to Sunday. Two, twice his or to Sunday and then Sunday again, and this can be a test period. Cool. So let's look at the average average traffic to that period per day was 11,000. I'm just going to have filled that out as a number of thousand 4 95.\nAnd what was the traffic after. And that's this red period here.\nOkay. So that's 9,961, as soon as the difference she wanted to take the 1,061 divided by the previous number and then minus one. And then we just kind of format that as a percentage, as a minus 13%. So we actually lost 13% of arch. Okay. So so that's interesting, right? It's not you know, not a huge drop and it might actually be worth it if that's although it was incremental, so let's work out how much you know, specific traffic we dropped, like how much traffic we would've got.\nIf if we had kept paying. So so if we had the average of the paid ads before and say how much traffic. Before on an average day for paid ads. So that was 6,000. I'm just gonna kind of copy this formatting here. So 6,166, that was how much traffic we we would normally get on an average day.\nAnd then we want to figure out basically what is the difference between a before and after? So are we going to take 11,405 minus 9,961? So we we, we lost 1500 clicks per day, so that was the part that was incremental. So so pay. In this case, if we were driving 6,000 paid clicks that, and only 1500 of them, it was incremental.\nThen the incrementality for that channel. About 25%. And that means, you know, if we were you know, previously spending you know, 10 cents a click then in reality, our true you know, cost per click was actually four times higher than that. You know, so 40 cents a click, we can figure out whether that's worth it to us or not.\n
What would be the best control and test periods to choose to determine the before and after effect of a change?
Upsert spends $3,000 per week on brand ads, what is their effective incremental cost per click?
What would be a good argument for keeping brand terms running even if it was ROI-negative?
Thanks so much for doing this\nI'll make the case to the CEO\nEven with low incrementality it looks like a no brainer to keep brand ads running\nEspecially if it means controlling our message\nI appreciate you!"