30 Sep 7 Ironclad (Slightly Fudgeable) Laws of PPC Campaign Structure
[This post is a companion to a panel session at SMX East in New York, Sep. 30, 2015: The Great PPC Account Structure Debate – #smx #23B]
It’s no secret that I’m no fan of needless busywork in PPC account management. Just as SEO firms used to bill clients for what Jill Whalen famously reduced to “boondoggles” — filling potholes that don’t exist, if you will — I’ve generally been wary of long to-do lists in PPC that may only negligibly affect performance, if at all.
Want proof? To take but a couple of examples:
- I wrote about Why I Hate Ad Extensions (tongue in cheek, of course!).
- I warned you about becoming addicted to negative keywords.
It’s in that spirit that I come to the Great PPC Account Structure Debate. Past contributions to similar debates have tended to showcase elaborate edifices rivaling the Great Pyramids. I sense that these more recent over-the-top structures will not stand the test of time and will not be viewed by future generations as Wonders of the World.
Quite simply, advocates of unnecessarily complex account structures underestimate the costs and drawbacks on several fronts: (i) transparency and collaboration when dealing with teams or leaving coherent structures for successors to follow; (ii) simple effort; (iii) cognitive load; (iv) de-aggregation of usefully aggregated statistics, potentially leading to a fruitless chase after more random outcomes; and (v) I’m sure there are some others I’m forgetting. The purported benefits are (usually) unlikely to outweigh the added complexity.
Past a certain point, our ability to set up and manage massively complex taxonomies is, well, taxed beyond the breaking point.
Google saw the handwriting on the wall long ago. Traditional, hand-built, fixed account taxonomies may continue to make perfect sense to a point, but think of just the following four AdWords products they’ve rolled out in recent years in attempt to save us from our own limitations in attempting to scale large-scale accounts:
- Product Listing Ads. While we still have the ability to control the metadata, group and bid in various ways, ultimately this system is meant to work on the principle of “load your whole inventory in, and let the system crunch the data in order to give users relevant results.”
- Dynamic Search Ads. Initially feared by many advertisers, DSA is now commonly used by advertisers with large sites as a “catch-all” for the long tail keywords and highly specific ad groups they just don’t have the time nor the wherewithal to set up. Similar to PLA’s, of course, control freaks have been given the opportunity to set up “targets” or groups, but regardless of how you use it, there is a chaotic, non-taxonomic, machine-learning quality to this product. Above all, in the pursuit of valuable incremental sales volume, the return on time spent here is more favorable than it would be setting up and managing that long tail campaign structure.
- Dynamic Remarketing. This form of remarketing for ecommerce typically outperforms regular remarketing. Rather than requiring a laborious effort to set things up in categories, this works off a feed and user behavior, showing relevant product images to users who’ve been considering just those products.
- Enhanced Campaigns. You can almost hear them thinking at Google: “Oh, no! There they go again, beating themselves up with laborious and overwrought campaign taxonomies. Let’s rebuild the architecture to allow geographic bid adjustments and mobile bid adjustments (to go along with time of day and week bid adjustments) to be laid overtop of any single campaign.” Hallelujah. (In some accounts, this would have had the effect of cutting sixty or more campaigns — all set up so that managers could control mobile and geo bids — down to one. Honestly, who can even imagine managing or reporting on sixty campaigns, multiplied by however many actual, honest-to-goodness campaigns are in the account? 1,800 campaigns, anyone?)
- Public service announcement: in providing that juicy new functionality to advertisers, Google removed bid control over the tablet channel. I have yet to meet a single advertiser that does not want this restored. Currently, Bing Ads allows up to a -20% bid adjustment on tablet bids.
Anyway, it’s time to get down to the 7 Ironclad Laws of PPC Account Structure. To fully focus on campaigns per se, that’s all I’ll address here. Ad groups are the next step down in the organizational hierarchy and in my mind are something like the fundamental building block of an account — so arguably, more important to understand and debate than campaigns — but let’s fully cover off campaigns, if we can, rather than getting bogged down in ad group stuff.
7 Ironclad (Somewhat Fudgeable) Laws of PPC Campaign Structure
- Law #1: The Fewest Campaigns Possible, Whenever Possible. All other Laws must answer to this Law.
- Law #2: RMKT, DSA, RLSA, PLA, Dyn RMKT, etc. require separate campaigns. This goes without saying.
- Law #3: Campaign settings provide a rationale for campaign structure. At this level, we can control a lot of settings, including mobile bid adjustments, geo bid adjustments, dayparting, ad rotation settings, bidding methods, ad rotation… you get the idea. Since it’s most convenient to control SiteLinks extensions at this level, a strong, thematic campaign structure can help control the look and feel of ads. This is an example of the many small exceptions and details that lead us to gradually move away (as appropriate, but no more) from our ultimate Law (#1). One more thing. If you rely on geo-specific ad copy in many geographic areas, you’re probably going to have to drift away from Law #1, and create many campaigns corresponding to those countries, regions, or cities. That decision shouldn’t be taken lightly, as you’ll be de-aggregating data and having a harder time testing ads, for example. Large companies with international divisions will also need to create more structure to account for different countries. It might even be cleaner to do this in multiple accounts, rather than using campaigns for that purpose.
- Law #4: Reporting requires narratives and stories that are (for better or for worse) siloed and segmented, such as talking about corporate divisions, product categories, or certain geographic regions. As such, these might be a rationale for campaign structures. (Overdone to the point of distraction, however, the practice might lead to overwork or excessive segmentation. And excessive segmentation can lead to both overwork and worse performance, given that tiny data segments are more random, so we’re more likely to make mistakes every time we take action in response to them.)
- Law #5: “Likes with likes” thematic control over campaign bidding matters, but only if the themes identify distinct behaviors that cluster, so that managing to the cluster (seasonally or otherwise) is an effective way of bidding and prioritizing more accurately.About that: there’s nothing wrong with using campaigns to pursue this efficiency while killing the “intuitive stories and helpful reporting” (Law #4) bird with the same stone, but you won’t be able to do everything with campaigns alone. Duh. Multiple attributes can be applied to any element in any smart scheme. AdWords is certainly no exception, allowing us to apply labels to any ad group or keyword. Labels might refer to high margin or high lifetime value keywords or ad groups; they might apply to hats and socks that people only wear (or buy) when the weather is frigid. Campaigns are a poor mechanism for handling the possibility of multiple dimensions impacting a keyword, whereas labels pretty much take care of it — and nothing says you can’t build your reporting around labelled keywords or ad groups.
- Law #6: Too many campaigns is a bad idea, because: (i) it hinders transparency and collaboration when dealing with teams or leaving coherent structures for successors to follow; (ii) it ramps up the required effort level more than people expect (think: these are boxes you need to look inside, that contain more boxes you need to look inside); (iii) cognitive load and/or sense of accountability (think: my desk is cluttered; I have to be ‘on top of’ 58 campaigns, but I never look past the top 4 and am nagged with a lingering sense of guilt that I say one thing and do another); (iv) de-aggregation of usefully aggregated statistics, potentially leading to a fruitless chase after more random outcomes; and (v) I’m sure there are some others I’m forgetting.
- Law #7: Too few campaigns can sometimes be a bad idea. This is especially so for large accounts with many brands or SKU’s, in which case serious serious thought needs to be given to Laws 4 and 5.
For many accounts (except for the bigger ones in the industry), the above laws should result in a reduction of the number of pointless campaigns. Many SMB’s get by with one or two key campaigns (suggested name: “Main”), with some additional ones to account for RMKT, and the various standalone products and channels that must be accounted for in separate campaigns.
So, if so many account managers are bent on heading down the path of needless complexity, the natural question to ask is: why? It comes down to two causes.
The first, without a doubt, is vanity. It afflicts writers (sorry), gardeners, and your great-aunt who is so convinced of her ability to arrange just the right amount of (way, way, too much, and utterly undustable) bric-a-brac on her many shelves, mantelpieces, and windowsills. Account managers want to create something that looks impressive, so someone can see what they’ve done. WYSIATI — so if what you see is impressive, you should be impressed. As a corollary to this, the structure becomes proprietary, and difficult to manage or undo. This is a natural inclination of anyone doing complex work, but it’s an impulse we need to avoid. If your work can’t get you paid without being camouflaged and set with booby-traps, you’re just delaying the inevitable (and burning bridges), IMO.
The second is a bit more innocuous, but deep-rooted: people’s approach to “organizing and storing” knowledge and information. Many of us are locked into certain approaches to information retrieval and storage, based on whatever corporate structure, era, or technology we came of age with/in. Take the whole idea of organizing your “desktop computer” into “folder structures.” Even I — who abandoned the practice as soon as he could — still follow such practices in certain areas. But with cloud computing, searchable everything, different ways of labeling things, etc., the old ways of doing things — which rather recently seemed very new in comparison with filing cabinets, paper files, etc. — are little short of quaint in many contexts.
Remember Yahoo? LookSmart? Dmoz? (Many readers, who have heard of Yahoo but not the others, now have no idea what I’m talking about.) According to these online information retrieval pioneers, information had to be categorized — placed in folders, if you will — by editors and experts. That continued in the tradition of library science with its Dewey Decimal and Library of Congress Classification systems. After pure keyword search took off, the categorization concept sort of melted away. We all watched as the confluence of available technology and user behavior led to more scalable information retrieval mechanisms. To be sure, something may have been lost in the process, but to most users beginning to type a keyword search into their phone, this new technology is more immediate. Archiving information in the proper places has no place in the average user’s world.
What on earth does that have to do with PPC category structure? Surely, companies need to segment and categorize information for the purpose of reporting and analysis, at least? Of course. Indeed, that should be part of the point: categorize intuitively and usefully, aggregate in ways that help in interpretation as well as improving campaign management, and ultimately, performance. Reasons that stray into “I just plain like putting things into boxes” are more suspect. And, I would argue, that (box-putters-into-fanatics) is the way a lot of people are hooked up.
New account managers, handed the task of setting up a best-practices and possibly boss-impressing campaign structure, may read a few posts and articles (and even a book) about account setup, and begin overthinking the task. Some simply fail the basic competence (IQ) test: they open multiple accounts, and use them like they should be using campaigns, aggregating them all in a super-account console commonly known as the MCC. High five! Perhaps, as they work to follow the many pieces of advice floating around, they worry about doing it wrong — so they do too much. Certainly a lot of pundits will be more than happy to foment that worry. And beyond that — back to vanity — once you’ve mastered a pretty sweet-looking PPC campaign structure (remind me again why this is something anyone should pat themselves on the back about?), why not keep on taking it to ever-more “advanced” levels, so someday you can out-pundit the pundits and get invited to SMX to speak about PPC account structure? This ain’t a scene, it’s a goddamned arms race!
Here it is: no one’s giving out prizes for the most impressive PPC campaign structure. Hands down, I’d be more impressed by the gourd that won second prize at the county fair. PPC account structures need to balance performance (that means the ability to invoke nimble control over campaigns, while managing even more directly to smaller segments such as keywords, plus, of course, practicing the granularity inherent in keyword advertising, to show relevant ads to searchers), reporting and narratives, reasonable aggregation of data, product and channel distinctiveness, and other common-sense rationales for categorization.
Beyond that, much of the magic happens in other places: with labels; in ad testing; at the ad group and keyword levels; and in the myriad levers PPC managers have at their disposal. Above all, the trick to campaign structure seems to be: don’t shoot yourself in the foot. And don’t create inscrutable, proprietary structures for the sake of your career. This is a team sport.