It’s common for consumers to click on an online ad, view a product and then postpone purchase until a couple of days later when they’ll go directly to the site and the conversion is made.

Furthermore, once an order is complete, that same customer could keep coming back to the site to make other purchases. And yet, measuring these instances of customer loyalty to optimise bids has often proven to be a problem for marketers.

The way latency and follow-on conversions are tracked can have a big impact on bidding decisions and revenue-on-ad-spend (ROAS) calculations. If you were to put all your campaigns on hold but conversions continued to come in attributed to Paid Search, how would you count it?

Popular analytics packages such as Omniture SiteCatalyst or Google Analytics use “Date of Conversion” attribution, which credits the order to the click on the day the purchase took place.

On the other hand, “Date of Click” attribution systems, such as Google AdWords Conversion Counter, increments revenue for the date of the original paid click, which may have been days or even weeks ago. And in both cases, because of latent and follow on conversions, it’s hard for marketers to react effectively because they may be working with incomplete data.

Consider the following sequence of events for a single visitor and how they will appear to the marketer:

 This presents a couple of challenges:

  • For many marketers, the first order is just a start. Ideally, you should be optimising bids, creative, and landing pages based on the expected lifetime value of a new customer. 
  • Waiting for follow on orders to be recorded over time means that bid optimisation will gradually push bids upwards, yet the keyword in question can become irrelevant to a seasonal or promotional campaign.  
  • How far should a marketer go in attributing customer loyalty to paid search? Merchandising and customer service teams will argue that they’ve created an experience which encouraged the customer to continue returning to the site, but should they get all the credit? 

    If a customer wouldn’t have found the website without going via a search engine then surely a search campaign must add value.

Smart Handling for Latency

You’ll see below some sample conversion latency data from a real online retailer. The green line points to first-time orders as a percent of their overall (one-year) total, charted by the length of time after a paid click. 87% of first-time orders take place on the same day as the click. 

The blue line measures resulting orders as a percent of eventual total subsequent orders. As the graph below demonstrates, while 95% of “first orders” happen within a week of the paid click, only 83% of subsequent orders occur during this time period. 

It’s this latency with follow on orders that causes problems for even the most meticulous of marketers.

Although optimisation and bidding processes need to take into account the latent revenues generated by a paid click, they also need to respond to market conditions. 

A number of advertisers deal with latency by disregarding the latest performance data and using a longer retrospective window to work out bids. This approach will simply increase the revenue attributed to a click when latent orders arrive days or weeks following the fact.

Although they are correct to do this, they are missing out on a chance to adjust bids quickly in reaction to market signals.

As an alternative to waiting for follow-on conversions to occur, marketers can estimate expected revenue from subsequent sales at the time of the first sale, and ensure maximum responsiveness to any alteration in conversion rates and purchasing patterns.

Using existing data to estimate the most likely value of a consumer from their first conversion will help marketers to make a more informed bidding decision at the time of the initial sale.

This is really very simple. If you know that a customer normally goes on to make two more purchases after their initial purchase then it makes more sense to bid for that first piece of custom.

Extensive analysis of historical data will give a marketer enough information on conversions to judge their average lifetime value. The most precise and responsive attribution and bidding systems use this expected lifetime value to determine bids, making suitable adjustments much closer to the date of click instead of the date of latent conversions.

The way a marketer can negotiate seasonal and promotional cycles is key to their success. With speed of the essence, estimating lifetime value allows for a quick response to changes in user behaviour, and will increase total revenue.

The ability to predict the impact of each consumer and factor it into bidding decisions allows the marketer to react to trends before competitors, which will increase revenues and ultimately market share.