The startup conundrum
Published: December 16, 2011
Author: Sean Marshall
Starting a brand-new PPC program has its challenges. Not having any history hurts on both the quality score front and the measurement side. Without a baseline, you really can’t tell how you’re doing. These pain points are exponentially worse when you are trying to sell a new mousetrap. Sure, you might know every relevant KW under the sun, but the general public has no concept of what it is you are trying to sell.
Search works because it allows marketers to capture user intent and adapt their offers accordingly. What do you do when your offer isn’t quite what people are expecting? Think of flash sale and daily deal sites one to two years ago. It’s phenomenal that you offer chairs and cabinets but when users looked for those things two years ago, they expected to see Ikea and Macy’s, not a site that may or may not offer it on any given day.
In new setups like these, learning is key. While testing and learning are critical parts of search, they aren’t nearly as dominant in old programs as they are in new programs. The beauty of an established PPC program is that there is a certain degree of predictability. In year 2 of any PPC program, you know what seasonal trends to expect around volume and consumer behavior. Year 1? Good luck.
Defining profitability and ROI targets are critical learnings that need to happen early in the life of a new search program. Over time, LTV and AOV for the search channel are defined, but you need data to make these determinations. One of the hardest things to do for an experienced SEM is letting go of hard ROI targets to enable learning. We’ve all done it on a small scale to test out a new product or new channels like YouTube, but when was the last time you did this for an entire business? It’s like kryptonite to a performance marketer. Sure, the channel’s performance is highly quantifiable, but it stings to pay any CPL when you have no idea what type of customer you are getting.
Search works best when targets are set for extended periods of time. That said, new programs require flexibility. Some boundaries should be set (breaking even is usually a good starting point) but targets should be adapted as you gather data. Sometimes this means weekly CPA target changes and that, too, is tricky to veteran SEMs. Elite search tech firms boast some of the smartest bid algorithms in the world, yet few, if any, can claim that their technology can hit a new CPA target in under three weeks.
It takes more than an algorithm to make these types of changes. In many cases, KW-level decisions are based on portfolio trends. If you’re expecting a 1% conversion rate in a new program and yet 80% of your portfolio has 20 clicks to date – what can you really do? How much data makes a trend? Hell, is this CVR even predictable? For an established e-tailer, sure, CVR should be predictable by category, but what about a case where your site and offering is constantly evolving? It’s nearly impossible to make assumptions about future performance if you’re dropping people on a completely different site from one day to the next. Some would say clean A/B testing is the solution, but startups can be finicky, and time sensitivity can lead to some interesting choices. (“Good morning, SEM guy. We completely changed the entire site last night. How’s conversion rate doing? Oh, and we probably deleted your tracking code in the process.”)
The nuts and bolts of SEM aren’t really any different, but the overriding strategy you adopt will be different. Take the case of a hot new startup with strong brand awareness. While I wouldn’t advocate doing this for established programs, excess margin generated by brand activity should be reinvested into non-brand efforts. This will greatly soften the blow to the company bottom line and give SEM a ton of wiggle room for testing. In cases like these, it’s OK to move away from break-even ROI and really push for profitable PPC – just at a portfolio level.
Of course, “break even” and “profitable” mean different things to different people. Targets should be rooted in your own set of economics rather than some arbitrary target. Implementing a solid attribution model is also a good way for a new program to get a leg up. Many would say that it’s best to get the basics out of the way first, but as the lines between different marketing channels are blurred, I’d say creating a multi-click/channel attribution model is the new basic.
Ultimately, every program needs boundaries. Startups need to be nimble and be prepared to adapt quickly, but just because you’re flush with cash doesn’t mean PPC shouldn’t be run the right way. Take some liberties and incorporate portfolio level goals. Let some keywords run hot to figure out what you’ve got, but don’t let it happen forever. Make decisions about your pages, but test them like you would in any other program. Don’t overreact to seasonal trends if you don’t even know what you’re getting into. Just be open to anything and make no assumptions. After all, you don’t have any data yet.
– Sean Marshall, Director of Search Engine Marketing
– Questions? Comments? Email us at blog at ppcassociates dot com.