We use cookies. You have options. Cookies help us keep the site running smoothly and inform some of our advertising, but if you’d like to make adjustments, you can visit our Cookie Notice page for more information.
We’d like to use cookies on your device. Cookies help us keep the site running smoothly and inform some of our advertising, but how we use them is entirely up to you. Accept our recommended settings or customise them to your wishes.

Anticipating the Flood: Holiday Bid Management Tips

As we approach what we hope will be a terrific holiday season online, it's important that ecommerce advertisers be ready to take full advantage of the opportunities presented. Adrienne Raynor detailed 6 growth opportunities a couple of weeks back, and I wanted to chime in with one more relating to bid management. Traffic spikes at the holidays for most B to C retailers and, by itself, this makes us all happy. But the fact that search volume increases does not mean that you can afford to spend more for the traffic. Let's say your "Bright Red Widgets" keyword generates $3/click in sales revenue on average. $1.50 goes to Cost-of-Goods, $0.30 goes to variable costs (credit card fees, pick/pack/ship, etc), $0.30 needs to go to overhead and profit contribution. This leaves $0.90 for marketing expenses, and if the CFO is willing to push it to the wall, you can have another $0.10/click to reflect some of the sales we can't track directly (or LTV, or some combination). You can profitably bid $1/click on "Bright Red Widgets". Beyond that, you spend money inefficiently; less than that, you leave top-line opportunity on the table. (These are oversimplifications but serve the purpose for now). Whether there are 5 clicks a day, or 5,000 clicks a day, the per click profitability turns south when the marketing cost to sales ratio exceeds 33.3% which means the profitability overall turns south as well. So, if during peak season, the search volume on "Bright Red Widgets" triples, that doesn't mean you can or should do anything differently with your bids. Indeed, if you decide to bid twice as much because the volume is high you will simply lose three times as much money as you would have by doing the same thing in off-peak season. But holiday tends to have a second effect: in addition to increased traffic volume, conversion rates do tend to increase -- significantly! {Sidebar: for most of our ecommerce clients average order size/value actually declines during gift giving season...possibly an interesting commentary on gift giving versus shopping for oneself :-)} This means retailers can and should bid more aggressively during the holidays. The key is to predict which bids to change, by how much, and when. We've written about this before, but it has been a couple of years, so here's a refresher:
  1. Historic data: The best predictor is historic paid search data. Looking week by week, how did the value per click change as we approached the holiday and thereafter? It is important to look at just the non-brand data, as brand search behavior is very different, and you're likely already at the top of the page on brand terms.Three different ways to view the data:
    • Dissociated: The worst option is to study value/click by taking the total value driven by search on a given day divided by the total number of clicks on that same day. The problem is that the value derived on that day may have been driven by clicks on a previous day. We want to evaluate the eventual performance of the clicks to anticipate bids.
    • Last click: Better, tie the value driven to the time of the last click rather than the time of the conversion event. This way we see what clicks on a given day eventually did for us.
    • First click: Better and perhaps Best, tie the value driven to the time of the first click within a reasonable window of time. Many many people simply hit the gas early hoping that they're accounting for this effect. We think measurement is a better approach.
    Here's what the non-brand Sales per Click ("SPC") data measured 3 different ways looks like for one of our clients which has a fairly high AOV and a relatively long click to order interval. Notice that there is a significant difference between tying the orders to the day they are placed, rather than tying them to the last click prior to the order. There is much less difference between last click and first click which is reassuring because it's not always clear how important the first click was to generating the next if they're both non-brand, and sometimes they reflect a completely separate shopping mission. It's a bit easier to see trends when we take out day-of-week effects by either rolling up data by week or by using a 7-day rolling average as we've done below.
  2. Study categories, not just aggregates. See if there are difference between major product categories in how the seasonal effects operate. React accordingly.
  3. Understand the calendar. Whether Thanksgiving is early or late has a big impact on seasonality, as does the day of week on which Christmas falls. Study the impact of shipping cutoff dates on performance.
  4. Factor in promotional effects. Promotional events seem to be driving folks to shop earlier and earlier.
  5. Be careful with day-of-week rules at the holiday. If you're day-parting, you might consider taking off those adjustments when we get to the big days. Black Friday is not like other Fridays, and Thursday 12/22 won't exhibit the normal Thursday periodicity.
  6. Avoid wasteful spending. Don't forget that post holiday, basic algorithms are going to be 'irrationally exuberant' about the performance data they see, and efficiency demands may suggest bidding much less than the algorithms would like to bid. We give this advice every year, but almost every ecommerce account we take over shows the same overbidding in the post-holiday period.
Applying a bit of thought and elbow grease to the problem of when to push and when to pull back is well worth the effort. If your paid search manager just says: "bid more, 'cause Christmas is coming!" ask them to do a bit more research. Happy Q4!
Join the Discussion