attribution

Effectively measuring the impact of display

What's the most effective approach to measuring the impact of display?

Most effective? Direct measurement. Clicks that lead to sales, even as part of a multi-touch path. Second most effective? Conversion correlation--you'll need a control group for this. If you're working with a display medium that allows you to specify an audience by unique id (say, email addresses into Facebook custom audiences), then you can split your list, and show ads to one group, and not the other. That's highly conclusive, too, and can account for view-through conversions without dowsing rods and tealeaves.

Have questions?  Drop me a line via the Contact form--I love to talk shop.

Up and to the right!

Moving beyond the 30 day last click attribution model

Why would a business use something other than a 30 day last click attribution model? What's the advantage?

There are two aspects to consider here.

The first, and by far the fuzzier, is the logical path to purchase. If someone doesn't visit your site for twenty-four days, are they really starting their shopping cycle over again, or continuing the original one?

Do they remember you at all?

Knowing your customers' shopping cycle will allow you to tailor this to your vertical. A car is a more considered purchase than a t-shirt.

The second is correlating those early clicks with purchases. Are very long sessions more likely to convert than short ones?

If so, then a longer window is important--once again, this is common on highly considered purchases. However, if your products are not a considered purchase, it's common for impulse verticals to be effective with attribution windows as short as one or two days. If the presence of older clicks doesn't correlate with higher conversions, then those clicks didn't contribute any value, and can be safely ignored.

Testing is recommended, however you'll find that most sites will do most of their business within 72 hours of the first click, so the fringe behavior of long sessions doesn't really move the needle. Some very successful sites have run unlimited windows, and still found value in them.

Have questions?  Drop me a line via the Contact form--I love to talk shop.

Up and to the right!

Methods of measuring the impact of paid search on other channels

What's the most effective methodology for measuring the impact of paid search on other digital channels?

There are two approaches--direct and indirect. Direct would be through the use of a Multi-Touch Attribution model. With this, you can directly see what other channels were supported by the participation of PPC. Indirect is a bit trickier--you need to be able to quarantine off a control group, which wouldn't be exposed to PPC. Since most PPC isn't user-level targeted, you might have to do this with something like Geography, or segments of your product catalog. Then, you could compare the performance of the PPC-included segment versus that of the PPC-excluded. If the difference is statistically significant, then PPC helped.

Have questions?  Drop me a line via the Contact form--I love to talk shop.

Up and to the right!

Including offline transactions into bidding decisions

How can we measure offline (phone, store) conversions and feed them back into bid decisions?

First, you need to get your phone orders into your attribution system. If you've built your own, then you can capture a code from the customer over the phone, which ties the CSR's browser session back to the customer's. Another way is to pick up clicks on emails (order receipts are a great example), where you know who they are, and can possibly track them back to purchases, post-facto. Once you have these orders accounted for in your attribution, then you simply include them in your attributed share coefficient.

Have questions?  Drop me a line via the Contact form--I love to talk shop.

Up and to the right!

Making bid management decisions attribution aware

How do we make our bid management decisions aware of the other touchpoints that led to a conversion (and not take all the credit)?

The simplest way is to do a channel-level coefficient. If AdWords is attributed 80% of revenue that it touches, then you can use a .8 coefficient in your math--either against the revenue that Google sees from within AdWords, or against the bids you're setting. More complex multi-touch attribution systems allow for different levels of control, but they can also become self-fulfilling prophecies--where you spend the most does the best, etc.

Have questions?  Drop me a line via the Contact form--I love to talk shop.

Up and to the right!