Improve paid search lead-gen with cohort analysis

When reporting paid search results, business to business (B2B) marketers often field a few recurring questions from stakeholders, clients, or internal teams:

“Why am I seeing an influx of search clicks and leads, but no opportunities or closes?”

“We spent $15K extra in search this month, so where are the results?”

“How does search contribute to retention?”

These are important questions because they are top of mind for stakeholders. However, the ability to strategically and correctly answer these questions about search campaign effectiveness over time requires both deep reporting capabilities and a strong grasp of your organization’s attribution model.

Many search pros are proficient in only one of these areas, often painting half of the paid search picture. The resulting gaps in measurement, analysis, optimization and how marketing dollars are spent leave stakeholders less than enamored with the results.

Search pros have other tools in the toolbox, though, that can better equip them to respond to those burning client questions.

Adoption of cohort analyses as part of paid search reporting can be a powerful means to assess trends, retention and path to purchase. It also allows for greater accuracy when analyzing campaign results over a dedicated window of time based on the time it takes users to move through the funnel.

In this post we’ll cover the basics of cohort analysis and how to deploy the model in your campaigns based on a lead-gen funnel:

Lead >  Prospect >  Opportunity > Customer

Defining and understanding a ‘cohort’

In marketing, the term “cohort” describes segments of users who share specific events or experiences within a specific time frame. Cohorts include purchasers, email subscribers, trial and/or demo downloads or any other conversion action in the funnel.

Regardless of segmentation, the value comes to life when monitoring these groups over time to analyze behavior throughout the sales cycle.

Without cohorts, lead-gen marketers are left guessing the “age” of customers in the funnel (how long they’ve been in the funnel). Marketers are then unable to find a true pulse of retention.

A typical analysis of paid search efforts involves looking at summary time frames and comparing them to the previous week, month or some other time period.

This is a great comparative tool, but it doesn’t solve a core issue: Looking at a summary time frame includes spend data that hasn’t had the time to produce a lead (or whatever funnel step we are viewing).

In other words, we are inflating our cost-per-lead figures by including spend that did not contribute to the leads we’re viewing. An example view of the click-to-customer path is pictured below.

One can find averages of performance based on fixed groups when comparing multiple time periods, but this approach doesn’t factor in any outliers. Whether it’s a group of repeat purchasers, cart abandoners or a subset that fades away during their journey, this mix of new and old customers inherently skews reporting outcomes.

If comparing the average pipeline per user from Black Friday year-over-year (YoY), average pipeline per user (APPU) may look fantastic as the exponential traffic drives pipeline, but what about the customers from last year’s Black Friday? Unless effective retention efforts are employed, those customers are likely declining in value, yet overall APPU reports are higher than ever.

Leaning solely on a metric like APPU can be dangerous in the long term because instead of taking account of the length of time customers have been in the funnel, it blankets revenue across the entire lifetime customer base.

Setting up the data flow

Shifting to a cohort model requires diligent up-front assessment and work; it’s crucial to ensure accurate data is being collected. The most important spreadsheet columns in this instance are the date and time stamps, such as “Original created date for the lead” and “Date when the lead transformed into its next stage” (think lead, opportunity, customer, date of first purchase and so on).

The dates allow measurement of the time it takes for users to move through the funnel and application of that knowledge to paid search reporting and insights.

Below is a list of ideal columns to have for an “opportunities” report:

• Lead created date.
• Opportunity created date.
• Lead ID.
• Source.
• Campaign.
• Term.

Time through the funnel

Once the right data is flowing and a statistically significant lookback window of results to review is available, it’s time to analyze the time it takes our users to pass through the sales funnel from paid search.

We want to understand the time required for an original lead to become a qualified lead, an opportunity and finally, a customer.

To set up a cohort analysis with ample data, generally, shoot for a six- to 12-month window of data. It’s vital to have a large enough date range so we don’t misinform paid search’s contribution to the marketing program.

We’ll start at the bottom to back into the results that we want to answer in the end:

• Days between lead and prospect.
• Days between lead and opportunity.
• Days between lead and customer.

Finding the date difference between lead and prospect is pretty straightforward. Take the prospect created date (the date when the lead turned into a prospect) and subtract the lead created date. Repeat for all leads, and be sure to exclude major outliers.

For lead to opportunity, it may be prudent to work in a separate document to avoid data confusion. Take the date the lead became an opportunity and subtract the original lead-created date. As you would expect, this date difference could be much longer than lead to prospect.

Repeat the process with customers.

After analysis, you’ll have a good idea of how long it takes for leads to move through each stage. You may even be shocked at how long the sales cycle is in a given instance. Right away you can tell why week-over-week views may not work well for some lead-generation campaigns based on how long it actually takes for progression through the funnel.

Selecting a percentile

The cohort model can be used to make faster and smarter search optimizations.

It’s not practical (or necessary) to wait for 100 percent of our leads to move through the funnel before making decisions. Choose the right percentile to use instead.

For example, taking the 75th percentile will help determine how many days it takes for the fastest 75 percent of our paid search leads to move through the funnel. This may significantly reduce the days between stages from previous analyses, but that’s OK. We know the rest of our leads will move to the next stage at some point. Remember, our goal is to make accurate decisions quickly.

Working within a shorter time window will require slightly higher cost-per-goals to account for the customers excluded from the model.

Another example: If the goal is $750 cost per customer and we’re working with the 75th percentile, we’ll want to increase that goal to $1,000. If we were to wait for all our customers to trickle through, we’d end up at a better cost per number than if we were just looking at the fastest 75 percent.

If the idea of working with a percentile sounds hard, remember that working in averages and a non-cohort model is already inaccurate. Our goal is to make actual optimization decisions with accurate data in as close to real time as possible.

Once a time frame and percentile have been defined, avoid including prospects, opportunities or customers that take longer than the dedicated time window to convert.

If the customer window is 30 days, and a customer takes 45 days to come in, including that customer in the 75th percentile window would artificially inflate the model numbers. These should live elsewhere in a summary table, not in the cohort decision-making model.

Developing reporting & reporting results

The key to developing accurate reporting is to ensure prospects, opportunities and customers aren’t being reported outside their time windows.

This means if a customer window is 30 days, we’re not viewing any customer results unless they’re 30 days old and have had that time to mature. To get an accurate cost per customer in this instance, we also want to exclude spend from the most recent 30 days.

We should only view spend in the maturity window for our customers or opportunities.

Following setup, the most accurate view will be available for volume, cost-per numbers and conversion rates through the funnel.

More than likely, it will come to light that performance is being under-reported because there are days of spend being accounted for, while lead volume hasn’t caught up yet. With the new view, you can begin to make decisions based on the most accurate data you’ve had the chance to work with.

In the above example, analysis has indicated that the fastest 75 percent of leads turn to customers within six months. Given this information, channel metrics can only be viewed for months one and two when analyzing cost per customer.

For cost per opportunity, channel metrics can be viewed for months one through five. Our leads can be analyzed in near real time.

Cohort analysis application for paid search

Forecasting. Understanding the flow and evolution of paid search cohorts in correlation to pipeline or revenue makes it much easier to forecast the behavior of a new subset of customers.

Retention strategy. Should you do more post-purchase? Comparing cohorts by day, week or month of acquisition by revenue generated from that group over the next six to 12 months will shine a light on purchase and engagement habit changes. If pipeline or repeat purchases don’t increase, it may be best to implement a retention or re-engagement strategy to guide users back to the sales journey.

Seasonality. Assessing date of first customer/purchase against repeat purchase or total pipeline will highlight users who fall off after a holiday or busy season. Using this data can help inform marketers whether they should double down post-season.

Geo-specific purchase behaviors. If employing international or geo-focused paid search initiatives, measuring revenue incurred month over month by location will make it clear where lifetime value (LTV) thrives or dives by region.

Analysis models vary greatly, and shifting to a cohort analysis or model can be a big decision. For many marketers, such a move is necessary for working with lead-gen campaigns.

Implementing cohort analysis into paid search reporting is often a powerful means of charting true long-term trends for retention, churn and attribution at a more granular level — and more importantly, bringing to light opportunities within paid search programs.