PPC has evolved significantly, and it is a good time for PPC professionals to discuss forecasting. Do experts do any type of forecasting for their clients or stakeholders? Do they use any tools for forecasting? This was discussed further by host Julie F. Bacchini during this week’s PPCChat session.
Q1: Do you do any type of forecasting for your clients or stakeholders? If so, what kind of forecasting and what industry or industries are you working in?
Forecasting has always been hard. But it is really hard now with the way platforms operate. @NeptuneMoon
A few clients want to get an idea around budget for the following year but that is the closest to a forecast that we do. We don’t like to plan that far out as forecasts are a waste of time I feel. @duanebrown
What would make it easier to do forecasting? @navahhopkins
Yes, and we are doing some now for next year. Some clients need them at various times of the year depending on when their fiscal year starts/ends. I work in a lot of industries (healthcare, B2B and B2C) so they all vary depending on their goals. It has been a lot more difficult over the past few years specifically. We have to consider people’s discretionary income, the platforms increasing the CPCs (CPC inflation essentially 4-6%). @Ichasse
Giving an idea of budget needed is one thing. You can project that, sort of, based on how CPCs increased in previous year. But telling them how many leads to expect to get from said budget is damn near impossible. @NeptuneMoon
@navahhopkins Someone else do the work for us. @duanebrown
Would you trust a number provided by an ad platform if one gave you an estimate based on budget and targets. @navahhopkins
@navahhopkins Honestly, probably more accurate planning tools would be helpful that take into account the increases in CPCs the platforms inevitably put in place and more accurate numbers when it comes to traffic/clicks vs. the generic we get now. @Ichasse
@navahhopkins I would use the platforms estimate as one available data point. @NeptuneMoon
Yes, a few clients i do estimates, just finished doing my final few for 2026 budget requests last month. i’ve seen people do this for pitches and i’m like Whaaaat with no data forecasting we can achieve X LOL and they wonder why the client isn’t happy after Q1…@Meriem
Everyone wants some omniscient view of their market. But the truth is it is always complicated and any forecast you provide is full of “it depends” @NeptuneMoon
I always present best case and worst case if asked – across industries – B2B/Ecom, etc. And it follows extra items which would be required in my opinion to achieve that target and it’s beyond ad spends. @alimehdimukadam
@navahhopkins If the numbers were used and fairly accurate 80% +, then yes. Most of the tools now are kind of garbage if I am being 100% honest. Most companies want to see growth each year for increases in spend, but the reality is we need a 4-6% increase in spend just to maintain current levels of revenue most of the time. You can see that is not a very good long-term formula to have in place, because eventually it won’t be worth the spend vs. that margins the company has. I have clients now that are asking “what other platforms” due to this. @Ichasse
Yes, we forecast. And we’re in quarterly revisions – holdco life. @JuliaVyse
For sure – it’s impossible to know with 100% certainty that x + y = z. The bigger issue is that not every brand is willing to pass conversion data in, so even if there was an accurate forecast, it wouldn’t be grounded in brand data unless there’s historical context. @navahhopkins
@alimehdimukadam – I do something similar when planning a new campaign. I create a ‘pro forma’ model that contains projected impressions, clicks, CPCs, CVRs, etc. that paints a potential snapshot of outcomes. It helps set expectations. @JeffreyHain
Most every client we work on requests forecasts. Mostly tied to our recommendations for budget based on performance previous month/year.
Cyclical / seasonal-dependent clients like travel/boat rental, we’ll usually have a year or two of historics to base our reccos on, but not always. Not always accurate, but the longer we have a client, the more accurate we’re able to predict these cycles. In the past month, here are a few forecasts I’ve ran:
Budget tied to ROAS forecasts (Google) for retail client. Basic revenue and ROAS projections by campaign type (pmax vs nonbrand search) – hockey/retail. CAC vs. value estimates (sometimes missing LTV)- toys/retail. Channel-level performance forecasts before agreeing on following months’ budget – travel, retail, professional services. @timmhalloran
We have a light prediction for clients, and a heavy one for internal. Planned vs spent down to the week per channel. All Hail Excel. @JuliaVyse
Hypothetically – if you were to be provided a forecast based on advertisers like you, would that meet client needs, or would they still want it based on their data even if they would hit conversion scarcity. @navahhopkins
@navahhopkins Benchmarking against other advertisers would be helpful. @NeptuneMoon
VERY helpful depending on how it’s done. (e.g. don’t compare me to a big beast please…) @Meriem
The issue is also, there is a point of diminishing returns, because platforms will give you the high quality users with high intent, but then when those users are exhausted, you end up with the lesser users so they can get that spend. Our goal is to find where that changes, but accounts never scale infinitely over time (unfortunately, lol). The platforms know when this is going to happen and a campaign’s performance (conversion) rate could end up being half what it was with the high intent audience it was sending traffic to at first. @Ichasse
Yeah – and to be clear there are no planned updates from Microsoft at this time around such a tool – just collecting input to go to bat for it if it would meet the need. @navahhopkins
Agreed! I’d find it helpful if it doesn’t fall into that “tyrrany of averages” trap that most of those generalized categories usually are written. Like it says, “Home and garden” which could mean a million different sub-niche verticals within. @timmhalloran
@navahhopkins Maybe… if we understood at a lowest level how those numbers were drawn up. @duanebrown
@navahhopkins – Clarity has some benchmark numbers. @alimehdimukadam
They do @alimehdimukadam but they’re not budget to leads/sales forecasting in the way this chat wants (at least I don’t think they are) @navahhopkins
Lots of clients want a forecast with no historical data to measure against! I use Google Trends to give some insight into Search trends by month over the last 5 years. @williamhboggs
Yup, can be a good reference point for people who want to run numbers with benchmarks. We always get this – what’s the industry benchmark? @alimehdimukadam
@williamhboggs have you tested Microsoft’s Audience Planner and if yes, how does it compare on helping with forecasting budgeting. @navahhopkins
Q2: What are client or stakeholders top requests of PPC forecasting? Do you find it difficult to provide what they want? If so, why?
Basically how much will it take to get x% more leads? And then what will the CPC and CPA be for those leads, etc. @NeptuneMoon
And it is very hard to predict those numbers. They also really, really want to know how they are doing relative to their competitors. @NeptuneMoon
What’s the industry benchmark? (Conv. Rate, AOV, CAC:LTV, ROAS, Churn, etc.) Some competition data – estimates from Similarweb with a disclaimer that it’s not accurate at all and only indicative. @alimehdimukadam
I always sound like a drug commercial when I present forecasts. We will have to test to see if this can be accomplished, but I cannot guarantee any results and these are only models based on past performance and what I believe will be the CPC increases over the next year. Like I mentioned in my other note, we do not know when the audience goes from high quality -> medium quality -> low quality in the tool. One may have a 5-6% conversion rate, the other 3-4% and the last may be <1%. Cost of acquisition will vary dramatically based on the quality of traffic the platforms are providing us with. I usually start with here we are now, here is what we will probably need to be spending to keep our same results and here is what we need to do for growth. @Ichasse
Results per budget, but a LOT of questions about opportunity. Take note Ginny and Navah: if your PMax performance builder tool only shows me potential for more conversions with no sense of cieling, it gets remnant budget. I need to tell big clients how much to spend in terms of how much can I spend, not just how much more conversions are possible. Benchmarking is a huge question. @JuliaVyse
What is the ceiling…. What is the breaking point…@Meriem
Spend vs conversions would be the biggest one for us… since we are 80% ecom brands. @duanebrown
I find it useful to put these conversations in the context of pragmatic math (i.e. can the budget deliver on paper based on auction prices and anticipated conversion rate). Clients usually are more coachable when they understand they’re embarking on an unwinnable math problem. @navahhopkins
One I always love is “how much could you spend” now i push that back in asking realistic figures, as ultimately I can spend as much as you want, depending on what you want to achieve lol. @Meriem
Honestly, it is like trying to predict the stock market, but clients think it’s somehow not. You can look at data, but that only gets you so far, and there are factors you can’t predict in play. @NeptuneMoon
@navahhopkins Have I got some client calls for you to join if you think my folks are coachable on math! Word to the wise: a brand advertiser does NOT know the definition of ‘can’t win’. @JuliaVyse
@NeptuneMoon I find the stock market much easier to predict than these platforms, lol. @Ichasse
lol some of my favorite client conversations are debates (respectful and civil) @navahhopkins
Yup, and I believe what gets missed here is brand building and other activities. Only performance marketing is a losing proposition over the long term. @alimehdimukadam
@Ichasse, what would make it easier to partner with platforms? Obviously, I’m interested in Microsoft, but general input is useful as well! @navahhopkins
@JuliaVyse Spot on, my clients do not like vague math or non-answers. A bunch of Alphas on a call, they will destroy you when you say nothing is predictable, etc… Those are the days where I need to go play some basketball or eat some chocolate after a call, lol. @Ichasse
@navahhopkins That list would be big! It is a tough one to answer in a short form to be honest. These companies who are not huge but mid-level and even small, see their investments as HUGE in these platforms, even though the platforms see them as tiny in most cases. Let’s be honest, the big brands spend more than 20 mid-tier and 200 smaller brands will spend combined. They don’t get the white glove treatment the larger brands get, who spend 500 mil+ on the platform. We usually get reps who are sales folks instead of actual account-minded folks, and the planners are not much help because their numbers never line up with reality once the rubber hits the road (not even close most of the time). @Ichasse
@Ichasse very valid feedback! @navahhopkins
Q3: What types of disclaimers or caveats do you provide with any forecasting for clients or stakeholders?
Life happens @navahhopkins
We can look at trends, but things happen that move markets in ways that trends won’t capture. Auctions are owned by platforms and can change on a dime. We strongly recommend a lot of brand cover to make noise during sales giving BOF channels the best chance. In cases where that’s not possible, we meet and communicate weekly/bi-weekly to stay on track and agree on tactical changes. @JuliaVyse
Basically, none of this is guaranteed and are offered for planning purposes only. Micro and macro factors can have a massive influence and cannot be foretold. Also, all of this assumes that client is not making any major changes to the way they do business. @NeptuneMoon
I always try to make clear that a projection is just that, a projection, and that if we could guarantee what was going to happen instead of making a guess at it, we’d be using those skills to make a lot more money than we are projecting out ppc campaign performance. @JoeWilliams
Oh god yes! If you ask for trends and planning on 30M and then only spend 10M, I’m not delivering to that plan! @JuliaVyse
Apart from standard disclaimers of no guarantee
I add areas which I feel they are missing and it could be email marketing, landing pages, referrals essentially all recommendations a head of growth would provide. Coz it all works in cohesion and not as simple as run ads, make money. @alimehdimukadam
We use past behavior whenever possible. If it is a down year economically, we will try and look at a year that was similar if we have enough data, but even that is very different since the campaign types are so different now. PMax and AI Search are very different than what we were running 5 years ago. For new clients who have never run ads, all we can do is use their current conversion rates we are seeing in their analytics platform for non-brand traffic to try and put together models. I am honest and tell them, these are models and in the financial world, they have Monte Carlo models that can forecast for extreme events happening, so you get here is the best case and here is the worst case and probably what reality is. I try and frame it like that for clients. @Ichasse
@NeptuneMoon yeah, no major changes like launching a new website where they change all their links, so you lose all your SEO value you had, do poor SEO on the site (which impacts PMax and AI Search campaigns)…. The best part is when they don’t tell you and the links for all your ads are broken one morning when you wake up! That has actually happened to me 3 times now in 10 years. @Ichasse
Q4: Do you use any tools for your forecasting? Do you like them? Do you wish they had capabilities they currently do not have?
We have proprietary tools, but we also use Reach Planner, Performance Planner, Environics, COOMB, Emarketer, Vividata, and others. oh, and sometimes we HAVE to use the planner in Meta. Join me in my hatred of that lying POS. @JuliaVyse
I’m super interested if ad platform forecasting tools currently make the grade and if they don’t, what about them falls short. @navahhopkins
Recent favourite to run with numbers – Claude Projects – add numbers with channels, no PII and create an artefact along with past trends and then do a future forecast. It works well if you have a couple of years of data and context. @alimehdimukadam
I wonder if Meridian or something similar can help with forecasting; haven’t explored it yet. @alimehdimukadam
I’ve been using GPT Project folders that I fill with client performance data, SOW details, project files, so it has the context behind the “ask”. Then I use Funnel. io or PBI to combine/normalize the data (if its needed), then take that CSV (and the comparison periods) and ask GPT to forecast. Most of the time, it’s pretty accurate. But occasionally, I’ll need to steer it because it will build averages from other averages and get crazy numbers. But it’ @timmhalloran
After doing this for many years as the person directly responsible for the P&L and now working with the folks who are responsible for the P&L I have gotten pretty decent at creating my own models and I have my own tools (glamorized spreadsheets) that I can put in data and adjust variables to match what their current results are to come pretty close most of the time. I have tested various tools and have not found any that were better than just doing the digging yourself into the brand’s data, mixed with some search volume information, and understanding diminishing returns as you scale. Pen to paper > any tool I have tested so far, basically. If I found a tool that I could trust, believe me, I would love to use it, because hours of my time would be saved each year! @Ichasse
I love this @Ichasse All the tools I have access to are just tools. It takes the collective brains of the smarties on my team to actually make something bespoke and powerful for our clients. @JuliaVyse
I find most brand mangers and CMOs are kind of lazy. They just come at you with a 20% increase in budget and expect a 20% increase in revenue, and they do not factor anything else into the equations. They have that data as well and can see how much CPCs have increase YOY, and they can see when you spend X and increased to XX over a period, the conversion rate either stayed the same or decreased. Unless they spend a lot on usability studies and then make changes to the site based on the user feedback, rarely do you see the conversion rates increase with increased spending. I actually called a CMO lazy once because they frustrated me, and that conversation went better than I expected it would, lol. It was probably my 4th call of the day, so my filter was breaking down! @Ichasse
@Ichasse Yup, and this cycle repeats because brands or companies for whatever reason do not implement suggestions, are looking for a silver bullet, and then end up changing the agency only to repeat the cycle endlessly. And some would change CMO, CMO would change agency – on loop. Everyone wants instant gratification….@alimehdimukadam
And to ignore the competitors in their markets. But we are so DIFFERENT from them @NeptuneMoon
Competitors are looked at only if their ad shows up on top of their own. @alimehdimukadam
My Good and lovely Google Sheets or Excel. I have a section for Google Ads specifically, Microsoft specifically and then i have Overall performance By Month for the last two years (Preferably) then i overlay a marketing calendar to predict things… LOL I’m so old school LOL – I wish people knew there marketing calendar for the next financial year before asking for forecasts LOL @Meriem
Q5: Are you using AI for any forecasting or reporting actual results vs. forecasted results? If so, what type of AI? And how is that working for you?
Not really. Our forecast vs actual doesn’t really need a lot of AI intervention, but I expect more forecasts to become more AI-powered in the near future. @JuliaVyse
If not….why not? @navahhopkins
Data privacy is a major concern. @NeptuneMoon
Privacy and compliance for gov clients. sheer cost for all other clients. these little bots are pricey! @JuliaVyse
I mean we kind of have to start integrating it into our work or be left behind by those who do over time. I ask a lot of market segment information. Size of market, what is the market cap, search volumes reported from at least 5 different sources, etc… I ask for the data in tables as I find the tools do a better job at not just spilling verbal diarrhea on the screen when you do that for data. They tend to give you paragraphs of text when you are just looking for a few numbers at times. I still use my own data as well, because I have worked in quite a few markets and have a pretty good understanding of each whether it is dental, hospitals, veterinary, home goods, retail, food brands, entertainment (streaming services), etc…@Ichasse
@JuliaVyse Yes, I don’t upload information into any of the tools because of this. If you have an NDA, DO NOT upload your client data into the tools unless you have your own computer with an AI tool loaded of your own to do analysis of data. They have already shown how the data gets leaked to the web. Even having the tools help with your email and such have allowed hackers to see every email you have ever sent or received. I only ask general stuff! @Ichasse
Most recently I’ve used: GPT – Project Folders, Custom GPTs (for more training built-in)
NotebookLM – I like the “mind map” function in it. And most recently it’s built-in some video recap features that allow you to pull in multiple documents and get synthesized replies on whatever you select.
Claude – The most recent one I’ve been experimenting with (like for 2 months now), I like the visualizations it’s able to product, and the connectors that come baked into the paid plan. Helpful for multiple use cases. Also, Google Forecast for ROAS predictions. A while back I did a correlation analysis in GPT to see the divergence between attributed-paid sources in the client’s backend database versus what Google’s forecaster tool told me the ROAS would be at a specific budget amount. We had what it had predicted for about 12 months versus what actually materialized (on Google at the end of the month, and within clients database via UTMs). And created a “Discrepancy Ratio” that we now apply to all the Google forecasts so it weighs it against real data to give an accurate number. So, if Google says I’ll get a 17:1 ROAS, I’ll add the ratio (it’s like 0.68) and then I’ll know that more realistically I’ll probably achieve a 10 or 11 ROAS. @timmhalloran
Claude Projects & Artifacts. Side quest – have it run on device with a model fine-tuned for privacy concerns. Gemini for Google Workspace does not train on your data so that is something people can use. (Read their ToS). My favorites are: Claude, Gemini, then comes ChatGPT. @alimehdimukadam
It came out this week that Gemini had more downloads on the app store than GPT, which was good news for Google and I wonder if some of the privacy stuff is part of the reason. @Ichasse
Ask you risk management resources about protecting yourself too – insurance and terms in your contracts regarding use of AI and responsibility and liability @NeptuneMoon
Claude is the best for coding currently, which is helpful for creating your own scripts if you play with the vibe coding. @Ichasse
Gemini overtook coz of Nano Banana. Their image generation model. Can’t believe anything you see online, very real. @alimehdimukadam
Question for AI users – do you double check the results? If so, how? And does that make it less efficient to use AI in this use case? @NeptuneMoon
200%. The trick is to not keep conversing in a single chat window for long. So if you want to do a forecast, break it down in 3-4 different chats – chunking. One chat window for one step. Take that result. And then start the second part. Have to double and triple check. @alimehdimukadam
And it can make some very stupid mistakes too. So you have to be the human in the loop. And there are some prompts that you can use and add in context which can prevent hallucinations and ask it to add context in the output like an assumptions log, etc @alimehdimukadam
@JuliaVyse I’d caveat that you should not all systems are created equal in this regard, but I agree that it needs to be a conversation, especially and absolutely with sensitive industries or data. 1) Don’t use the “free” version if you want to run analysis or ask strategic questions that give access to proprietary info – overtly or otherwise. Just not a good idea. 2) Turn off data sharing options in those platforms if they are turned on (the subscription levels we use don’t train on the data we give it). 3) Be transparent with clients on the ways that you use AI and the ways that you don’t. I’ve found that most assume it’s being used, so being clear and intentional about explaining how ends up building trust versus degrading it. 4) I think there’s value in working with legal to build a type of generative AI snippet into your SOWs but I won’t speak on that piece, since I’m not a lawyer. I know what we do, and what I’d recommend, but Googling ad law blogs may serve you better on this piece. @timmhalloran
@NeptuneMoon Yes, I’ll check my results manually if I do any analysis. If you start to get comfortable with not checking, that’s when you’ll create a problem for yourself. This is most obvious with copywriting. But once you get into analysis, the numbers are harder to corroborate, so just stick to manually checking the math, and keep an eye on any “weighted average” or “grand total” columns, especially if they include averages like CTR, CVR, ROAS etc. @timmhalloran
I use AI as my co-working buddie, so if i’m like hmmm i wonder, i’ll have a chat with AI, we chat about trends and any new competitors within the market and new articles about certain things… etc etc…. I think i’m going to name my AI buddie.. I shall call it “Marvin the Paranoid Android” @Meriem
PPCChat Participants
- Julie F Bacchini @NeptuneMoon
- Navah Hopkins @navahhopkins
- Ali Mehdi Mukadam @alimehdimukadam
- Boggs @williamhboggs
- Lawrance Chase @Ichasse
- Duane Brown @duanebrown
- Meriem @Meriem
- Julia Vyse @JuliaVyse
- Jeffrey Hain @JeffreyHain
- Timothy Halloran @timmhalloran
- Boggs @williamhboggs
- Joseph Williams @JoeWilliams
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.