Today’s PPCChat discussion theme was timely, given the numerous changes across PPC platforms. PPC professionals must be prepared to clearly explain these shifts to their clients and stakeholders. Here is the screencap of the entire session hosted by Julie F. Bacchini.
Q1: How do you think about learning periods on the various platforms these days? Does it vary by platform?
There is always a learning period! I find it varies more by volume than platform. High volume campaigns will learn faster. @beyondthepaid
You definitely have to plan for it in search campaigns now, which historically has not been the case. I find myself automatically sort of discounting the first 4-6 weeks of a new campaign as just volatile. @NeptuneMoon
Up to a week usually. Depends on how fast the data comes in. There is learning on both sides – Platform & PPC Professional. @alimehdimukadam
There’s always a learning period, but I think it penalizes you more in Meta and potentially a little bit in Search campaigns versus some other types. @revaminkoff
It’s a relatively new behaviour for search. my clients are still surprised by it. and it really plays hell with 2wk campaigns. @JuliaVyse
At least Meta has been like this since the beginning though. Google wasn’t and now very much is. @NeptuneMoon
I feel with newer platforms like Reddit, learning period may only be a day. @AmberDeedler
It’s making short-duration campaigns very hard to do / actively working against them arguably. @revaminkoff
I can only imagine how hard it is to do a short blitz campaign now. @NeptuneMoon
its’ really rough. @JuliaVyse
I start with manual cpc on search to accelerate it, and then switch it to any of the other strategies after the campaign has 15-30+ conversions. @alimehdimukadam
Yes, especially with Prime Day. @AmberDeedler
I’m a max clicks girlie. It works, particularly in LTO world. Try our new burger, RIGHT NOW. @JuliaVyse
Yeah time sensitive, short duration campaigns you’re pretty much stuck with max clicks or manual. @beyondthepaid
For anyone in Canada who appreciates this, the Masala Burger at A&W is now a regular menu item. giver a try! @JuliaVyse
It does vary by platform, I try to avoid creating a new campaign or make substantial changes. Sometimes it’s totally unpredictable, I made a change 2 months ago in a campaign, and it recovered in 10 days, so made another change 1 month ago and the campaign never recovered from it :S @ernohorvath
@JuliaVyse that ad is making a lot of noise on X coz of the language. @alimehdimukadam
Yes, it does seem to vary by platform, and it’s super important to manage expectations with clients. I once encountered an issue at a real estate agency where a client was angry about their results after 30 days. I told them that the problem wasn’t the results it was not effectively managing the clients expectations with the learning periods, build optimizations, etc. @DiiPooler
Q2: How do you talk about learning periods on the various platforms these days with clients/stakeholders? Does it vary by platform?
There have been a lot of questions around this posted here in the last few months which is why I wanted to do this topic today. @NeptuneMoon
We have to remind clients frequently that we need to let campaigns run for a couple of weeks without major changes, and that performance will vary widely before settling in. Clients who have done search for a while struggle with this. @beyondthepaid
It varies by platform, but I warn them that ALL platforms have some kind of learning phase where the machine will evaluate the audience v message. Some of them overspend first, some underspend first, all of them will be weird on day 1-5 @JuliaVyse
My short answer is that clear expectation setting is even more critical if you’re doing search advertising these days. The era of turn it on and it fires right up and shows you what to expect from performance are over. Now, you have to expect a minimum of 4 weeks of the machines trying things no matter your bid strategy. @NeptuneMoon
This is particularly important for clients that have been doing search ads for a while. Their expectations do not match today’s reality. @NeptuneMoon
I always tell clients that the first 4-6 weeks will be the most expensive. @revaminkoff
I was working with ChatGPT/Claude last week to generate exactly this. Setting client expectations and sharing a timeline of what to expect when – Week 1 to Week 6+ @alimehdimukadam
Clients are also not used to changes putting search campaigns back into learning either. @NeptuneMoon
Yep, “the first month is the worst month” is what I always tell them. @beyondthepaid
But I think part of the problem with learning periods is that they don’t align with clients’ strategic goals or marketing tactics – especially when things are happening to existing campaigns, or there’s a new initiative that needs to be pushed. @revaminkoff
The change thing has been a rollercoaster. Flag for any highly regulated industries/public sector: that learning period after a change WILL trigger another ad policy review. uuuuugh @JuliaVyse
We can tell clients about learning periods, but that doesn’t change the business realities @revaminkoff
@revaminkoff 100% true. The algo assumes everything is a long game. @beyondthepaid
My best bet has been evergreen search terms for fast food with ad changes during LTOs. I don’t even bother with product names for short burst campaigns. @JuliaVyse
Google has done a terrible job of communicating this too. The way they talk about campaigns still makes it seem like you can hit the ground running efficiently. But nope. That is not how it works. They could help on this front by being honest about the early days of campaigns more overtly. @NeptuneMoon
For short burst public safety campaigns like wildfires that pivot into evacs or more comms, we just do our best and try to loop the reps in. Shout-out to Kyle, our rep for Dentsu Canada. He’s VERY pragmatic and good at communicating on these topics without promising the impossible. @JuliaVyse
I think part of the problem is also not every campaign is built to or designed to scale to infinity. But the system is looking for that. It is a disconnect. @NeptuneMoon
Agreed – I always feel bad for clients with volume capacities. where they can only handle so many leads or sales @revaminkoff
Hyper targeting isn’t right for GM, just to keep it in the family, but it can be right for a lot of B2B and specialized products. @JuliaVyse
@JuliaVyse enterprise reps are worth their weight in gold. Clients tend to listen better to Google reps than to us haha. @beyondthepaid
And the recommendations of increase budget don’t help either….@alimehdimukadam
I think another issue, speaking as someone who has worked lead gen PPC forever, is that we used to be able to figure out workarounds or ways to get an e-commerce first ad platform to work reasonably well for us. But those options are pretty much gone now. So it also creates a greater divide in how campaigns function and optimize across e-commerce and everyone else. @NeptuneMoon
The issue is, it destroys trust. Well, I created this campaign for you, dear client and I cannot do anything in the next 2 weeks and let’s just hope for the best. It’s hard to communicate your ‘edge’ against the other agencies/freelancers who just do the same. Or maybe I’m just a bit negative today. @ernohorvath
I hear you @ernohorvath it does create a trust issue, or at least opens up potential thoughts of “what is this person/agency actually even doing for me” that you really don’t want especially if it is a new client you’re still building trust & rapport with! @NeptuneMoon
Well, and we’re the experts. so do we know Google or not? Unannounced or poorly communicated changes like that make us look bad. @JuliaVyse
@NeptuneMoon Yes, this is mainly an issue for new clients with new accounts. @ernohorvath
It’s extra hard if something you executed for a client in the past just cannot be replicated today. They don’t understand it and when you explain that Google changed the way things work, some don’t want to hear that. It sometimes reminds me of the sign in the beauty salon my mom went to when I was a kid that said “I’m a beautician, not a magician” cause I feel that way about search campaigns some days @NeptuneMoon
It’s also hard when the client has understandable business urgency – like a construction client we work with who is looking for people who want to remodel/rebuild their decks @revaminkoff
Q3: What are your strategies to deal with learning periods on the ad platforms? Does it vary by platform and/or type of advertiser?
We try to put a moratorium on any major changes in new campaigns for the first 2 weeks. @beyondthepaid
I like to build it into my pacing sheets and expect a front-loaded budget. No changes right after launch unless it’s major/error correction. And I try to build visibility into my traffic sheets. Like, this will launch on x date, learning period ends on x date+5 days. And we put it into the first mid-flight comms “now that learning periods are ended..” in other words, communicate more is my strategy, rather than tech workaround @JuliaVyse
I have been including more detail in what to expect for the first month, including language that basically says “we are monitoring everything but changes will be made very judiciously in the first 4-6 weeks of any new campaign due to the impact of ad platforms learning periods.” @NeptuneMoon
Only client communication. Platforms will keep changing, tweaking and testing all their AI’s more often than not now. @alimehdimukadam
For us it’s about bake rates for conversions. 30 days is way too long for high-investment learning. Ideally, secure a testing budget and push back if you know you will have data anomalies in the first 2-3 weeks. Try to hold hasty decisions at bay. I would add it varies by subchannel as well Shopping vs text vs Pmax etc. Not just platform @MariaCorcoran
I generally will add negatives in the first month, but not much else. I am watching asset and ad performance, but not touching anything unless it is truly something that must be changed. I will also adjust bidding if the campaign can’t get traction. @NeptuneMoon
Once a campaign is established, if a client wants big changes, I will also send documentation about the campaigns potentially going back into another learning period, so there are no surprises and I’m covered when/if it happens. @NeptuneMoon
Another strategy is to layer in launches based on investment increases. Start with top 5 DMA in brand and non-brand text in Google, next week to Bing, 3 weeks expand to top 10 DMA, add in shopping, 2 more weeks add in another 10-15 DMA, adjust based on performance, then add in PMAX. Layer high CVR efforts with things like NonBrand to control CPA. It offsets the pain of learning. @MariaCorcoran
I’ve found that consistency in my messaging to the clients to be the biggest hurdle. I’ll set the expectation that learning will take 3+ days to a few weeks at the outset, and that we should treat that window of time differently than the overall campaign’s performance. We’ll give aggregate totals regardless (1 mo., 2 weeks etc.) but also notate when a campaign’s learning ended, so when we cut that timeframe out, they’ll see the true CPL or ROAS is higher – sort of like treating that portion of spend as experimental budget. @timmhalloran
But this is where that has failed in the past. As we know in marketing, consistency is essentially “all things being equal”. Ad experiments, measuring KPIs by repeatable time frames; if you don’t remain consistent, you’ll get a reputation (perceived or real), that you cherry-pick data by how it looks. So, back to the learning periods, if a client immediately has success even within the learning period, it’s easy to slip the bounds of that experimental timeframe and just say “We started doing great right out of the gate” so we won’t split the measurement. But then you’ve set up a precedent that that’s the new expectation next time learning comes around. @timmhalloran
Set the expectation that the first 4-6 weeks of any campaign will be the most expensive. But also don’t let the learning period keep you from doing what you need to do – I think it sounds more intimidating than it is. @revaminkoff
@timmhalloran all of this! exactly this. @revaminkoff true, but also Narcan. but also eating disorders. But also wildfires. It isn’t ‘that’ intimidating, but it can and will cause a cascade of other delays/spend issues. @JuliaVyse
Great points @timmhalloran I think we have to always have the talk about learning periods and I find it helpful to also say something along the lines of “Google will move in many different directions and test/try many different things in the first weeks of a campaign. Sometimes we get lucky and it hits a homerun that keeps delivering results quickly. More likely though, it will follow a multitude of paths as it tries to find the ones that will consistently produce the conversions we are asking it to find.” @NeptuneMoon
I typically like to account for the learning period, but launching campaigns on Mondays rather than Thursdays as an example. Explaining that it’s a collaborative approach, especially if it’s a new campaign, we have to give the campaign data & watch how it utilizes it. The first 30-45 days is a learning period. @DiiPooler
Q4: What are clients/stakeholder’s biggest misconceptions regarding learning periods on ad platforms right now? Does it vary by platform?
Most clients do not think learning periods exist. @revaminkoff
That search is still instant. That learning periods always inevitably mean low spend at first and not high spend to learn on. That learning periods are optional. that learning periods are the same on all platforms. @JuliaVyse
They also underestimate the fact that it’s very hard to get out of the eddy if you’re in it.@revaminkoff
Not caring about hitting stat sig because they want to be in market making money asap.@MariaCorcoran
Clients don’t understand learning periods at all. @beyondthepaid
Even clients with their own algos!!! @JuliaVyse
That platforms will absolutely chase the easiest conversions until there are no more before trying to capture better (often harder) ones. @NeptuneMoon
That they can see their ads immediately on all platforms after the campaign is live and can search for the keywords to be greeted by a nice ad advertising their business @alimehdimukadam
What you choose as your conversion action impacts everything, including learning periods. @NeptuneMoon
@alimehdimukadam funny, sad, and true. @beyondthepaid
Meta is always instant especially with retargeting. Google not so much. @alimehdimukadam
As I said in another answer, clients who have been doing search advertising are often thrown by the whole idea of learning periods for search ads. @NeptuneMoon
@NeptuneMoon I mean yes! And on top of the easy conversions, hopefully, you won’t have a stakeholder who cares what budget is driving specific purchases vs benefiting the client holistically. @MariaCorcoran
Not to toot my own horn, but my clients think I can make Google – or anyone – do things they wouldn’t usually do. NEVER pull rabbits out of hats, clients will think you can do it daily. @JuliaVyse
@MariaCorcoran I feel like I am having more “what are we ACTUALLY trying to do here with this advertising?” conversations because of this now. @NeptuneMoon
That it’s a simple formula. X amount of days + X amount of spend = X amount of learning period
“Why are we still in learning” “Because…reasons.”It used to be fairly measurable (15 conversions in a 7 day period, etc.), but now it’s “We see most success if you have X amount over an X timeframe, but learning periods vary based on other factors included but not limited to: audience size, spend, conv, geo, time frame, and 981 other signals.” So there have been instances where I give a timeframe, only for it to take longer. What I’d be curious from all of you, is if you have a formula that’s more often than not, a good way to gauge it outside of system-prompted explanations (that seem to change daily). @timmhalloran
Also, when campaigns don’t want to serve. Or when they drop off a cliff for no real apparent reason. That’s also super fun! @NeptuneMoon
At this point I almost never share formulas. it’s fine to use them for benchmarking, but it’s too risky to set expectations. @JuliaVyse
When campaigns don’t want to serve or perform very erratically in the beginning, it’s not necessarily something that can be resolved in an hour. @revaminkoff
I think we have to use more language that is “generally” or “often we have seen” rather than this is what will happen within 4 weeks because we just don’t know. @NeptuneMoon
I feel a blog post coming on: When the Learning Period is longer than your campaign. @NeptuneMoon maybe ppcchat dot com would like such a thing? @JuliaVyse
I think I had a conversation like this with a client once on why it stopped working and we have to duplicate or hard reset audience signals and then turn it back on: “The machine learning thought it figured out the audience (and probably did) but then it milked it dry and couldn’t figure out where to look next – like a hamster wheel, it couldn’t see past historical patterns, so we need to pause it, and restart things to help it “think outside the box -allow it room to test and iterate again”‘ @timmhalloran
Yep – jiggle the handle – it often works @beyondthepaid
I think too, it is important to step back and think about it like this. We have a general idea of how long learning periods should be, I was just asking myself what I can learn on $Xk in 3 weeks for non brand in a highly competitive space. Well I know what I need to spend and conversion volume I need to drive. What investment level would I feel comfortable as a loss for learning at, if it were my money? @MariaCorcoran
Platforms are heavily using a Blackbox called AI – the reason it’s called a blackbox is coz they themselves don’t know how these Terminators will perform. We try our best to be Arnold and once the campaign is live – it’s a fight till the movie ends. And yes sometimes the best move is to use only one dialogue and get paid for it – I’ll be back….to check the stats another day. @alimehdimukadam
I know we will never get an answer on this, but I really, really want to know how much the AI uses competitor data to make decisions in my accounts. I suspect it is a lot more than most of us would be comfortable with…@NeptuneMoon
All I know is increased budgets, even when they say you are tapped out, does not equate to increased conversions, let alone in the first few weeks. Having insight into how they allocate that spend, saying we can spend more for more conversions, is total BS. They just spend for the sake of “learning”. Stopping my rant. @MariaCorcoran
Q5: What could the platforms do a better job of when it comes to both managing learning periods and helping us to explain them to clients/stakeholders?
Be more forthcoming about what their machines do. Some platforms spend very lightly at first, others spend a lot and then normalize. You can say what to look for without revealing the secret sauce. @JuliaVyse
More transparency in general would be helpful. Even if we could just reference documentation from a platform that outlines how they approach a new campaign. It may spend double your budget for a month with minimal conversions as it figures out which visitors will convert. It may spend your whole budget (and then some) on one click one day that does not even convert. I’d like more insight into how it latches onto things too and what it takes to get it to move in other directions. @NeptuneMoon
And how it works per campaign type. A reach campaign needs to touch everyone in the audience in a certain time. While a ROAS campaign needs to learn about the potential audience and evaluate. I expect these learning periods to be very different, and that should help us communicate to clients. @JuliaVyse
Similar to what Julia, Julie said, a recurring message I see across Meta, Google and others is “You’re restricting your performance. Don’t use your custom audiences. Instead, do nothing and let [Advantage+] do it all for you.” But it doesn’t give context into why. If it explained why, or what audience/intent signals I’m limiting by using 1P audience data to help guide its decision, then at least I’d have context into choosing one over the other. Even a “war game” gameification type thing where you can see how limiting certain targeting types turns off certain signals vs. keeping them on. Right now, I can’t visualize it without spending a chunk of change on an A/B test. I’ve done enough experiments that prove the opposite. So stop playing. @timmhalloran
@timmhalloran too right! In fact I think Google does a better job of this than Meta. They push for 1st party in my experience, explain why and how it works really well. @JuliaVyse
I haven’t completely read through this yet, came across in a research. It’s Google’s patent for an autonomous campaign optimizer. https://patents.google.com/patent/US20190139079A1/en @alimehdimukadam
Smart campaigns just got smarter? @JuliaVyse
Agreed with that. I understand how signals work in Google better than Meta for sure. @timmhalloran
Also, stop assuming that every advertiser both has more money to spend and wants to spend more money. Plenty want to see some results before they want to spend more. Many others are capped at that spend level for the time being. Every answer cannot be “well just remove bid limits” or “RAISE BUDGETS.” @NeptuneMoon
Shall I take you through the budget appropriation process, Google???? @JuliaVyse
Also, did a large campaign build last night in Meta. The AMOUNT of times it would turn on AI enhancements or Adv. features every time I duplicated a build to use it as a base for a new audience, was TOO D*MN HIGH. Like it’s literally resorted to sneaking AI into ads or ad sets after the fact. Conditions me to not trust it even more than I already do. @timmhalloran
Offline. editor. now. @JuliaVyse
Unicorns and rainbows lol. I’ve been wanting that since forever. @timmhalloran
AI being shoved in our faces is on my last nerve. @NeptuneMoon
PPCChat Participants
- Melissa L Mackey @beyondthepaid
- Julie F Bacchini @NeptuneMoon
- Ali Mehdi Mukadam @alimehdimukadam
- Julia Vyse @JuliaVyse
- Reva Minkoff @revaminkoff
- Amber Deedler @AmberDeedler
- Erno Horvath @ernohorvath
- Dii Pooler @DiiPooler
- Maria Corcoran @MariaCorcoran
Related Links
Stop the wasted ad spend. Get more conversions from the same ad budget.
Our customers save over $16 Million per year on Google and Amazon Ads.