How To Predict AdWords Spend Many Months In Advance?

[ergophobe]

Thoughts on Predicting a Paid Search Budget?

I'm trying to figure out a good way to determine the proper paid search budget month to month looking into the future.

I know the short answer: if your ROI is positive, keep spending. The problem for that is when proposing an annual budget and planning for June while still in January, I need to be able to make some estimate of a logical amount to spend based on ROI, search availability, etc.

In my case, there are certain branded terms where the ROAS will always be great, but the search availability is too low to sell at capacity. Meanwhile, generic terms are plentiful, but if we push bids too high, the ROAS is bad and so that becomes a problem. So what to do?

I've thought of a few possibilities, but none of them are satisfying (and BTW, I'm really new to AdWords, so these are just a few ideas that have crossed my mind).

1. Lost Impression Share method

In this method, I look at lost impression share due to budget, figure out how much more budget it would take to reach 100% impression share and voila! I have my number
--- future budget = (current budget) + (extra impressions * CTR * CPC)

The problem is that our IS lost to budget is small, especially for terms with conversions. So to increase impression share, we would need to bid higher, and that would reduce ROAS significantly.

Resources:

Estimating necessary budget increase to reduce impression share lost to budget: [searchenginewatch.com...]
Risk of pushing bids to increase impression share lost to rank: [searchenginepeople.com...]

2. Increased rank method

- Look at keywords that are "below first page bid" and have good ROAS and have lost impression share due to rank.
- Figure out how much you would have to bid to get X conversions
- Find the marginal cost per conversion
- When cost per conversion reaches the threshhold where it's no long profitable, that's your max.
- Use that bid in the KW planner to estimate total budget.

Problems
- Time consuming to get all ad groups added to KW Planner... is there a way to automate that?
- Assumes that conversion rate is consistent, but that may not be true. As you saturate people with more ads, you may be getting people less likely to convert.
- Not a single one of our terms that is below the first page bid has a single conversion.... so is there any point in bidding on them?

3. Cost per conversion focus
- filter for keywords with more than 5 conversions in the last 30 days
- sort by Cost per Conversion.
- See what Impression Share is and see if there's room to grow.
- See what happens if you put these KWs in the KW planner at higher bids - how much does another click or conversion cost?

Problem
- we only have 3 such terms and their impression share is over 95% for two of them and 72% for another.
- KW Planner says that no matter how high you bid on the 72% term, you don't get more impressions.

4. Number of conversions focus

- 1. figure out how many conversions we want
- 2. figure out how many impressions we need for that many conversions
- 3. use the keyword planner to determine what bid level is likely to get us to that level of impressions (a bit of a guess because the keyword planner does not account for our quality score)
- 4. plug in those bids and arrive at a budget.

This is a lot of work using the methods I know. Is there a quick way to do that and would it yield the kind of info I want?

If you have any ideas/methods/insights... please say!

[werty]

Not a single one of our terms that is below the first page bid has a single conversion.... so is there any point in bidding on them?

I often find that even though I am below the first page bid I will still show up on the first page as well as get impressions and clicks. So for me this number is usually BS and I do not trust it. I tend to bid what makes sense from a CPA standpoint. If a keyword is a dog and not making sales bidding it up may not get you sales, however it may cost you way way more money. So I am careful around that... especially when it is my own money.

If you are doing PPC for this entire company/client I would ask for 10-15% budget increase to allocate to Bing, and setup the account (30 minutes), import your Google campaign (5 minutes) lower bid prices as needed across the account (could take a long while) and look like a hero(priceless) for the new sales it drives.

I try and find out the target CPA and then go backwards. If they have a target CPA your goal then becomes to drive as many leads/sales/etc to meet/beat that CPA.
If certain terms generate leads/sales worth more per click then you may need adgroup or campaign CPA targets, but I think your methods above all sort of spell out madness.

Find a target CPA and then drive whatever your budget allows and experiment to fill the void between actual CPA and target CPA. Example if you can get them to sign off on a 10k a month budget and 50$ a lead, that is 200 leads you need to deliver. If you are paying 30 a lead now and only driving 150, that leaves 5500 (55% of your budget) to fart around and play with to try and get the remaining 50 leads. As you get a better handle on your experiments you can reduce the bidding/strategy to only target positive returns and build that into your bucket for the following month. It allows you to test while still delivering value.

I am selective with whom I will work with, but a baseline for me is that they must know what they are willing to pay per lead or sale, either as a % of sales or CPA on a lead. And I try and educate them that if there is a positive return you keep growing.

[ergophobe]

>>your methods above all sort of spell out madness.

:-)

Maybe this is something that just can't be done the way I'd like, since it's all about predicting the future and I guess I'd be retired on my stock portfolio if I could do that.

I should explain, this has little to nothing to do with optimizing for performance within the constraints of a given budget on a day-in, day-out basis. This is entirely an exercise in long-range, ballpark budget planning (estimating budget 2-14 out). So when you say

Find a target CPA and then drive whatever your budget allows

That's sort of the quandry. The method you mention is basically how it's done now - you have a 10K budget and then optimize for that. Spend less to earn a sale than it costs via other channels, and everyone is happy.

But where does the 10K come from? This is a highly seasonal industry where product expires. So if you have inventory in July and don't sell it, you can't make up for it and sell it in August and you can't just grow spend month over month (I would imagine this would be similar to the problem retailers face transitioning from Dec to Jan).

The problem is that budget is set way in advance in the context of other media buys - print, banner, radio, PR budget - and there's always just sort of a seat of the pants guess. We're sitting in November saying, "Okay, we'll spend $10K on paid search and $10K on print and $10K on radio $5K on PR in July."

I can say, "Let's budget $30K for paid search in April," but let's say April rolls around and a week into it, it's clear we're not going to spend $30K at an acceptable ROAS. We can dial that back, but now it's now far too late to go an buy a print ad for April or sweeten the deal for partners/affiliates... and meanwhile April inventory expires and either doesn't get sold or sells for closeout pricing just to get something for it.

I imagine to at least some degree this is the same as concert promoters and Ticketmaster, airlines and Expedia, restaurants and Open Table, an appliance repair service and Angie's List....

>>number is usually BS

Check. Thanks

>>10-15% budget increase to allocate to Bing

We already do about 10-20% on Bing depending on how it's performing.

[eWhisper]

It sounds like you have an account that has quite a bit of history; for accounts like that - this is what you can do if you want a 'quick & easy' type of method:

Export your spend by month for the past 3-5 years.
Chart out how your spend changes by month and your average growth rate YoY.
Look at the projections for search growth (or use a trending system like Google trends, keyword tools, etc).
Determine if there are other channels you'll use this year that you didn't (remarking, GSP, etc) in the past.

Now use your past years spend, multiplied by the projected search growth or your YoY trend, add additional channels as necessary and there's a 'quick & dirty number'.

Trending past years data is usually a better indication than trying to go through impression share as IS is based upon a lot of factors and trying to get IS by month by some reasonable CPA estimate is a lot of work as you've mentioned -so trending can be more easier.

The issue is that you have some months where you'll spend more/less than projected - no projection is perfect. So you can either find other channels that provide decent ROAS (maybe not perfect) such as facebook, admarketplace, gemini, etc and move budgets as necessary. The other option is to just tell your boss this is the best you can do and your goal is to be within 10% of projections based upon historical data.

[ergophobe]

Thanks eWhisper. Taking on paid search oversight (not management by any means, but just giving direction to contractors) is totally new for me. I've pretty much hardly dealt with it before.

I think I still struggle with the fact that though the account has history, I don't know whether the spend ever made sense. I can tell it wasn't too high, because ROAS was acceptable. But I struggle with figuring out retroactively if spend should have been higher.

>> history
Yes, the account does have a fair bit of history.

>>goal is to be within 10% of projections

That is more than adequate. Even on a month-to-month basis we only strive to be within 10%. For the longer range projections that is more than close enough.

[RhinoFish]

But also realize that others in the auction can wreck your predictions, by changing their own behavior. :-)

[buckworks]

>> but now it's now far too late to go an buy a print ad for April or sweeten the deal for partners/affiliates

Too late for print ads, yes, but strong affiliates could boost your exposure within a matter of hours if you bribed them adequately. ;)

Are you running ads on the content network?
Are you building remarketing lists? Mailing lists?

[ergophobe]

>>by changing their own behavior. :-)

You have no idea... the next year will probably result in huge changes there which, you are right, presents a challenge.

>>strong affiliates could boost your exposure within a matter of hours

Some yes, some no. But yes, we can do things there. It's usually not a matter of bribing the partners - they have long-term negotiated contracts for percentage of revenue - but a matter of lower prices. Lower prices improve conversions, so at a certain price break, they will "feature" us. But keep in mind, that can be a lot more expensive than just raising bids. I think a 20% discount is the absolute minimum that gets partners to respond. Though it isn't within hours, it is within days.

In most cases our contracts bind us to price parity with all partners, but we do have the ability to break that for "members only" deals that require the user to be logged in and therefore do not show the price publicly. So we can limit the exposure of a deal somewhat.

>>Are you running ads on the content network?

Yes, though just paused Bing display network b/c of poor performance lately.

>>Are you building remarketing lists?

Oh yes. Funny story - Mrs Ergophobe and I served as models for a major ad (full-page in print in a several places, banners, a lot of remarketing). I started having the feeling I was stalking myself. Everywhere I went, there I was! Fortunately we were looking away from the camera so at least I wasn't constantly confronted by myself looking back at me from all sorts of news sites.

>>Mailing lists?

Yes. Both ones we've built and ones we've bought.

And you didn't ask, but...

>>Flash sales
Those too

>>direct media buys on partner sites?
Yep

>>Listings on relevant directory sites etc?
Yep

>>Direct mail?
Yep

>>Conference presence?
Yup

>>Hosted events for corporate buyers?
Yup

>>Blogger events, Instameets, "influencer" marketing?
That too

>>dancing bear sign spinners at busy intersections?
Mmm... no. Probably not a good fit for us.

Generally speaking, paid search is the most flexible, reactive method we use, so it's likely to fluctuate in real time by a fair bit, but I am always at a loss at the beginning of each month and especially at the beginning of each year with coming up with a logical paid search spend number.

Ultimately, we tend to look at same month last year and make a seat-of-the-pants stab at whether to raise or lower that. Somewhat like eWhisper suggests, but I feel not quite "scientifically" enough. It always feels like a guess because I still just don't really know how to decide whether last year's number was the right one.

[buckworks]

Mrs Ergophobe and I

A lovely couple! :)

Ultimately, we tend to look at same month last year and make a seat-of-the-pants stab at whether to raise or lower that.

That's kinda what we do, too. The CEO's pants actually.

just paused Bing display network b/c of poor performance lately

Just wondering, what were your expectations there and how were you defining success/failure?

[ergophobe]

>>A lovely couple!

Why thank you! Keep your eyes open.... you may see it soon ;-)

>>expectations
Hmmm our expectations there were

- ROAS that would result in a positive return
- ROAS inline with what we see at similar properties (I'm not involved in those accounts, but I have access to all the data and it's all the same industry, so very useful).

So basically compared to the search network and compared to other properties on the display network, it had poor CTR, terrible conversion rate, very high cost per conversion. Bing search has traditionally outperformed Google search, but Google display network dramatically outperformed Bing (like by multiples - triple digit percentages not double-digit).

Too many things on my plate to dig in and try to figure out just now, so we paused it until we can watch it more closely and try to identify the source of the poor performance.

[ogletree]

One thing I like to look at when setting a budget is understanding how each day of the week and each month trends differently. Most campaigns have predictable impressions for days of the week and month. Most people see that Monday and Tuesday have huge impressions and Friday is way less and weekends are even less. Obviously this is not the same for everybody but is what I see most the time. If you set a hard daily budget for every day and every month you are going to miss out on some things. It might makes since to have a huge daily budget for Monday and a small one for Friday. December might be the worst time or your best time. You might just shut your account down on holidays or you might need to increase your budget for Holidays. Like eWhisper said it really helps to look at past performance. Set up a budget for each month and even for each day. You can set up a AW script to change your budget if you have it planned out. Make sure you have it set to send you an email each time it is changed so you can make sure it happens.

[ergophobe]

We do have good data for several years in the aggregate. A succession of contractors who all think the previous one didn't know what he was doing keep changing campaigns and ad groups around so it's hard to track at that level.

But we tend to look at same month last year and the year before when looking at this because it's a highly seasonal industry. Variation within the week is much less pronounced.

>>AW script

Ah! Didn't even know there was such a thing. That's a treasure trove... not really my job to deal with it on that nitty gritty level, but that looks worth some serious investigation.

[ogletree]

Let me know if you have any questions about AW scripts. I use them all the time. Most of the time you can just find a script already written and just make a few changes.

[ergophobe]

Thanks ogletree. Currently swamped with other priorities, but I'm trying to get as much automated as I can so that I'm less frequently swamped with other priorities. Hope to revisit soon.