{"id":15767,"date":"2021-04-14T11:46:03","date_gmt":"2021-04-14T06:16:03","guid":{"rendered":"https:\/\/www.karooya.com\/blog\/?p=15767"},"modified":"2021-04-14T11:46:04","modified_gmt":"2021-04-14T06:16:04","slug":"ppcchat-ppc-forecasting","status":"publish","type":"post","link":"https:\/\/www.karooya.com\/blog\/ppcchat-ppc-forecasting\/","title":{"rendered":"PPCChat | PPC Forecasting | 13th April 2021"},"content":{"rendered":"\n

Julie F Bacchini<\/a> has hosted this week’s PPCChat session with guest host Richard Fergie<\/a>. The session was focused on what kind of forecasting PPCers are doing? what is the hardest part of doing forecasting? any resources which are helping experts to ease out this task? and more.<\/p>\n\n\n\n\n\n\n\n

Q1: Are you currently doing any kind of PPC forecasting? If so, what kinds of forecasting are you doing and how are you doing it? How confident do you feel about your forecasting?<\/strong><\/h4>\n\n\n\n
<\/div>\n\n\n\n

I’m doing quite a lot of type 3 forecasts at the moment – mainly around trying to figure out what will happen after lockdowns etc. And I’m talking to a few people around scenario planning with media budgets. @RichardFergie<\/a><\/p>\n\n\n\n

“It Depends” #PPCChat<\/a> Wait, does that work for all PPC questions like it does for all SEO questions in #SEOchat<\/a> ? @WTFSEO<\/a><\/p>\n\n\n\n

I think everyone wants some insight into “what happens AFTER” these days… Regular forecasting is more along the lines of “what should we spend” or how will results change if we change our budgets. I can’t say I always feel super confident in forecasts… @NeptuneMoon<\/a><\/p>\n\n\n\n

Youtube forecasting through Reach Planner; appears to be more accurate for large budgets. @PPCKenChang<\/a><\/p>\n\n\n\n

I do a lot of Type 2 forecasting. If we increase our budget, how will that affect the results. I’m not super confident in the forecasting. It’s usually an educated guess and then just testing to see if guess is correct. @AllisonMiriani<\/a><\/p>\n\n\n\n

We do a significant amount of model development + forecasting — most of it is a time-series or causal, though we’ve started integrating more qualitative data into both of those to see if we can’t increase robustness. @DigitalSamIAm<\/a><\/p>\n\n\n\n

1 Probably going to soon be looking into the type 2. About to work with some clients that are already doing PPC and will start with analysing whether they need to be doing more.. @TheMArketingAnu<\/a><\/p>\n\n\n\n

I do normal KPI’s formula based forecasting with help of data studio. It’s fairly accurate. @bufoting<\/a><\/p>\n\n\n\n

Not right now because the account I’m working in has other priorities. As for confidence in forecasting, I’m confident I can move things in the right direction, but how far is IMO unpredictable. @stevegibsonppc<\/a><\/p>\n\n\n\n

We are doing some of type 2 forecasting, going by Richard’s categories. Typical example is getting new product & estimating how reallocating budget will affect results. @heyglenns<\/a><\/p>\n\n\n\n

Yes, especially for larger spend accounts. We look for growth opportunities and try to estimate diminishing returns as we expand out from core campaigns. @AndrewCMiller<\/a><\/p>\n\n\n\n

I use the various forecasts you can pull via @Kenshoo<\/a>. I also use what you can pull from @GoogleAds<\/a> as well. The accuracy of the forecast provided by Google seems to be lesser than that of Kenshoo for whatever reason (in my experience) @BrettBodofsky<\/a><\/p>\n\n\n\n

Every god damn day. Largely done by combining insights from IS lost\/gained, seasonality, and macro economic factors (ie obscure holiday, stimulus checks, unemployment, etc). @JonKagan<\/a><\/p>\n\n\n\n

(alternate answer): Generally, apart from “Is this worth testing?” I’m not too concerned with forecasting. For me, the goal is improvement. How much improvement is something that you’ll soon discover. @stevegibsonppc<\/a><\/p>\n\n\n\n

Right now doing a lot of Type 1 and Type 3 forecasting. Getting new clients that have either not tried google ads before or don’t have much data. For the type 3, we are looking at 2019 data to project & throwing out 2020.@selley2134<\/a> <\/p>\n\n\n\n

1) As an avid analyst and optimist, I strongly dislike forecasting. I once forecasted taking a brand from $20K in monthly sales to $60K within 3 months. Found a great marketing angle and hit $500K in the 1st month. So, I don’t do too much of it. @yaelconsulting<\/a><\/p>\n\n\n\n

<\/div>\n\n\n\n

Q2: If you are doing PPC forecasting, are you using any tools in your process?<\/strong><\/h4>\n\n\n\n
<\/div>\n\n\n\n

Excel was always my go to. Not sure if it still exists – but there are actually some predictive formulas in there that could be used for forecasting. @TheMArketingAnu<\/a><\/p>\n\n\n\n

We use several – ranging from Excel \/ Tableau (for the basic ones) to R (for more complex models). We don’t use a ton of platform-based forecasting models, just because their accuracy has been suspect and it’s easier to integrate other data into your own model. @DigitalSamIAm<\/a><\/p>\n\n\n\n

I use the tools built into the ad platforms. So for Google Ads, I would use the Budget Simulator or the Campaign Target CPA Simulator. @AllisonMiriani<\/a><\/p>\n\n\n\n

Using an internal tool to help with seasonality forecasts. For brand new forecasting, I use SEMRush & the keyword planner to get ideas of cost & potential impressions. (keyword planner has been underestimating CPCs recently from what I have seen) @selley2134<\/a><\/p>\n\n\n\n

As mentioned in my previous tweet. When I have access to Kenshoo I\u2019ll use it to help with forecasting in some instances. I am interested to learn a bit more about @RichardFergie<\/a>\u2019s @ForecastForge<\/a> @BrettBodofsky<\/a><\/p>\n\n\n\n

I have a few tools in my toolkit (as you might imagine!). R is a great programming language for experimenting with lots of different models. And I also eat my own dogfood with @ForecastForge<\/a> – particularly when I’m trying to make something that others can edit. @RichardFergie<\/a><\/p>\n\n\n\n

1: the commonality for these models is that all produce a range of outcomes + a confidence interval for the range. I strongly dislike point forecasts, because (1) there’s a degree of luck involved; (2) there’s volatility in everything and (3) the message matters. @selley2134<\/a><\/p>\n\n\n\n

A bit of google trends and @spyfu<\/a> and @semrush<\/a> @JonKagan<\/a><\/p>\n\n\n\n

<\/div>\n\n\n\n

Q3: What do clients typically request when it comes to PPC forecasting? Do you think that what they ask for is what they actually need? Do they ask for things that are difficult to forecast?<\/strong><\/h4>\n\n\n\n
<\/div>\n\n\n\n

Once on board, clients have better expectations I think of what can be forecast accurately. The prospects who are interested in working with you but haven’t yet want hard numbers of how much do they need to spend to get x number of leads. @AllisonMiriani<\/a><\/p>\n\n\n\n

Typically clients want to know what competitors are doing, how much they should spend, and what return they can expect. …and they HATE that all I can give is estimates. @JuliaVyse<\/a><\/p>\n\n\n\n

On my end they tend to want to know with x amount of dollars how much revenue do we believe we will be able to generate and at what ROAS. What I find more difficult to forecast is new initiatives with little to no historical data. @BrettBodofsky<\/a><\/p>\n\n\n\n

Clients often want absolute figures for “how much do we need to spend on this platform to get X conversions\/sales\/leads”. SO MANY VARIABLES, especially with new or newer clients. I always look at the post click process as part of any forecast too…@NeptuneMoon<\/a><\/p>\n\n\n\n

I’ve yet to meet the client who doesn’t want to know the answer to “if I give you X, what can I expect in return + over what time period?” Of course, that question is maddeningly difficult to answer. @DigitalSamIAm<\/a><\/p>\n\n\n\n

Potential clients (during sales process) are usually the ones that press for performance forecasts. I usually refuse to forecast performance outside traffic potential. @selley2134<\/a><\/p>\n\n\n\n

1: Because if there are site issues, my work could be brilliant and executed perfectly, but the end results (conversions\/sales\/leads) will be lackluster. @NeptuneMoon<\/a><\/p>\n\n\n\n

Generally clients will want a MoM growth projection which involves KPI’s. I have found some clients getting into the details of even asking for forecasting of metrics like imp, CTR & then comparing with actual figures. That was too much work. @bufoting<\/a><\/p>\n\n\n\n

Will echo @AllisonMiriani<\/a> here. Tough to please client with forecast answers & don’t often bring them ‘in the tent’ unless they know how to contextualize forecast info. @heyglenns<\/a><\/p>\n\n\n\n

Luckily for me our clients don\u2019t ask for exact perfect forecasts and if anything this year should teach us models don\u2019t have certainty – as we tried to model Covid etc.. @runnerkik<\/a><\/p>\n\n\n\n

“I have no idea what our seasonality is, do you?” @JonKagan<\/a><\/p>\n\n\n\n

<\/div>\n\n\n\n

Q4: What is the hardest part of doing PPC forecasting for you? Do you find any one type to be more challenging (new initiative\/platform, predicting performance on existing platforms w\/ budget change or general trends)?<\/strong><\/h4>\n\n\n\n
<\/div>\n\n\n\n

Seasonality and new announcements. no matter the tool, I can’t predict whether people will like your new stuff. I can expect a certain increase in searches when I run awareness media at the same time, but expect a high bounce rate because it’s brand new. @JuliaVyse<\/a><\/p>\n\n\n\n

I definitely think new platforms\/new clients are the hardest. If client is in same\/similar industry to current clients, it’s much easier.@AllisonMiriani<\/a><\/p>\n\n\n\n

Type 1 – forecasting for activity you haven’t done before has been my most furstrating. Just relying on Google data and being frustrated when Google says your CPC should be \u00a31.50 – but when i go live, i’m efficiently running at much less than that level! @TheMArketingAnu<\/a><\/p>\n\n\n\n

Anything with a small sample size can get dicey quickly; distinguishing between linear + exponential growth early in a series is tough; and determining where diminishing returns + market saturation hits is rough. @DigitalSamIAm<\/a><\/p>\n\n\n\n

New Accounts are the most difficult for me. Really have to rely on outside tools & your knowledge of those tools & how they skew. If there is reliable data in the account it makes it much easier for me to give reliable forecasts given a change to budget or trends. @selley2134<\/a><\/p>\n\n\n\n

The hardest part about forecasting for me is being confident in something that I deep down am not confident in – I get asked often how much should we spend and with no account history I can give a guess at best. @runnerkik<\/a><\/p>\n\n\n\n

For B2B it’s how wildly #s<\/a> range because of variables, esp post-click ones Julie mentioned. Going from CPC to page conv rate, to demo\/MQL, to SQL, to Sale; a 10% swing on one of those #s<\/a> messes up the end goal # entirely. @heyglenns<\/a><\/p>\n\n\n\n

New platform forecasting can be tricky. I like to follow the advice of @duanebrown<\/a> and install platform pixels on sites way before we might actually start advertising on them to gather audience data. @NeptuneMoon<\/a><\/p>\n\n\n\n

Testing new platforms. so many clients want a firm plan or even a guarantee from a test. it’s a test! part of testing is managing the risk that it might not work perfectly. @JuliaVyse<\/a><\/p>\n\n\n\n

Forecasting is tough when 1. Keywords have low search volume 2. There’s less or no historic data 3. Industry is very niche 4. Sales cycle is long or complex 5. Attribution is not clear. @bufoting<\/a><\/p>\n\n\n\n

It is half science\/math\/historical data, and half “I need to put in a buffer for civil unrest, politics, insurrection, recession”. That 2nd part gets a bit tricky. @JonKagan<\/a><\/p>\n\n\n\n

Performance can be very sensitive to changes. Our goal is to maximize a client’s efficiency – maximize client results with minimum client input (ie. budget). One needs to make sure that the forecast doenst become a limitation to how high to aim. Tricky stuff. @yaelconsulting<\/a><\/p>\n\n\n\n

<\/div>\n\n\n\n

Q5: When you provide forecasting, how do you position it? Do you provide any caveats or disclaimers when you share a forecast? If so, what do you typically say?<\/strong><\/h4>\n\n\n\n
<\/div>\n\n\n\n

Yes, always! position your data based on what is known, what you’re trying to accomplish and what you can’t foresee. Never provide a forecast without a position imo. @JuliaVyse<\/a><\/p>\n\n\n\n

I provide more disclaimers than a Lipitor commercial. In all seriousness, a forecast is an educated guess, based on limited data and specific, fixed assumptions. There is luck and randomness everywhere that we simply can’t account for in any projection. @DigitalSamIAm<\/a><\/p>\n\n\n\n

That it is an estimate and only a best guess estimate, not the actual performance of the actual campaign; actual performance could be much better as I tend to be conservative in my campaign forecasts. @PPCKenChang<\/a><\/p>\n\n\n\n

I always lead with “these are all estimates based on the available data and are in no way a guarantee of what will happen if we do X.” Will also callout any particularly volatile elements in play that could erupt at any time and make the forecast useless. @NeptuneMoon<\/a><\/p>\n\n\n\n

1: also, every forecast\/projection includes a range of outcomes + a probability that the actual falls within that range (a confidence interval). Life is weird and people are weirder. @DigitalSamIAm<\/a><\/p>\n\n\n\n

So many caveats. “this is just to give us an idea”, “these tools are good for trends but not absolute numbers” ^these are what I typically use in the sales process. For my own clients I don’t have to say this as much. @selley2134<\/a><\/p>\n\n\n\n

I definitely give caveats! It depends on: – competition – market place – weather – Google updates – above the line activity (tv, outdoor etc) And some other things that my magic eye just can’t see. @TheMArketingAnu<\/a><\/p>\n\n\n\n

Caveat: Attribution Lag If the client happens to be looking in-platform () I like to make them aware that it\u2019s possible that not all conversions have been reported yet. So the data they are looking at may actually look better at a later time as things backfill. @BrettBodofsky<\/a><\/p>\n\n\n\n

Clients ask for an upper range & a lower range for a media plan I share. I usually give a disclaimer that actual numbers may vary because I don’t control the ad platforms. @bufoting<\/a><\/p>\n\n\n\n

Every forecasting document I do has 3 pages of caveats\/T&C’s. It may just happen to be in white ink and the background of the slide… @JonKagan<\/a><\/p>\n\n\n\n

I say, “These are conservative numbers and here’s why I think we can hit them.” @stevegibsonppc<\/a><\/p>\n\n\n\n

<\/div>\n\n\n\n

Q6: What do you wish you had to make PPC forecasting easier?<\/strong><\/h4>\n\n\n\n
<\/div>\n\n\n\n

Time to play around with @ForecastForge<\/a> @DigitalSamIAm<\/a><\/p>\n\n\n\n

Based on other users earlier responses, it sounds like I should brush up on R. @BrettBodofsky<\/a><\/p>\n\n\n\n

Just some acceptance. 2020 changed all consumer behaviour worldwide, and the algos still don’t know what to do. They have 20+ years of history, and one wild year of bonkers behaviour. We’re in testing right now fam. we’ll see. @JuliaVyse<\/a><\/p>\n\n\n\n

Knowing that wishes can be unrealistic, I’d like data science skills in-house and clients who readily agree to pay for them. @heyglenns<\/a><\/p>\n\n\n\n

Some stability. As others have said the last 13 months or so has just been crazy which makes forecasting forward insanely hard. @selley2134<\/a><\/p>\n\n\n\n

More industry verticals data, only because I\u2019m incorporating predictive formulas in my estimate forecasting. @PPCKenChang<\/a><\/p>\n\n\n\n

I wish that the forecasts within the platforms were better. It would save having to figure out ways to do your own forecasting if you could actually reasonably rely on what you see in platform as a decent starting point. @NeptuneMoon<\/a><\/p>\n\n\n\n

If forecasts came with confidence intervals so we could explain the clients the probability of this event happening, it’d be awesome. @bufoting<\/a><\/p>\n\n\n\n

<\/div>\n\n\n\n

Q7: Are there any resources you\u2019ve found that have helped you with your PPC forecasting or your thinking about PPC forecasting?<\/strong><\/h4>\n\n\n\n
<\/div>\n\n\n\n

This is the perfect time to plug my PPC Forecasting article that … I haven’t written yet. @RichardFergie<\/a><\/p>\n\n\n\n

Rather than tech, I’ll give @siliconvallaeys<\/a> book a shoutout. “Digital Marketing in an AI World” gives good eg’s along lines Richard has been talking. Extols power of blended data. @heyglenns<\/a><\/p>\n\n\n\n

Excited to see the answers here. @selley2134<\/a><\/p>\n\n\n\n

Very big on answer the public for awareness search. Looking at what questions people are asking is a very good way to start answering them. and not always on search! answer them on pinterest if it’s a planning question for example. @JuliaVyse<\/a><\/p>\n\n\n\n

Anne Cushing\u2019s book Making Data Sexy. @PPCKenChang<\/a><\/p>\n\n\n\n

Welp as usual I\u2019ve learned a thing or two just from attending this #ppcchat<\/a> sesh. Certainly have some homework to do following this. Always so great to hear insights from fellow ppcers. @BrettBodofsky<\/a><\/p>\n\n\n\n

I’ve found the below video by @michellemsem<\/a> to be very useful to learn about forecasting. youtu.be\/a-wwDDnks8U<\/a> @bufoting<\/a><\/p>\n\n\n\n

<\/div>\n\n\n\n

PPCChat Participants<\/strong><\/p>\n\n\n\n