Five Steps to Predicting High-Value Customers Based on Their First Order

Feb, 2 2017

AdobeStock_71191350.jpeg

Dear marketer,


You spent blood, sweat and tears (and a big chunk of your budget) to acquire new customers. Congratulations, your acquisition stats are up! But just when you are ready to rest on your laurels you are beset by a recurring nightmare: Attack of the 1-time buyers! "Why aren't they returning for a second purchase??" you fret and make a bold resolution: "I'm gonna win them back no matter what it takes!"

You launch a barrage of win-back campaigns with irresistible offers, and some of them are now coming back. You're a hero! But there is a cloud raining on your parade: a pesky analyst from finance is reporting a drop in margin, AOV and LTV. She shows you data that these customers you lured back made small orders, maximized discount rates and never came back - a net loss for your company.

What if you had a magical crystal ball to tell you which customers are most likely to be of high value over the long run based solely on their first purchase behavior? You could target your win-back campaigns to attract THOSE customers! How? By looking at older customer cohorts and seeing which first-order attributes yielded the most valuable customers post first order. Below is a step-by-step recipe. If you have AgilOne's platform you can pull all the metrics directly. If not, give that analyst from finance a call so she can query your company's data to provide you the needed numbers.

 

Step 1: Select your cohort


Select a customer cohort that (a) has had enough time to return for a second purchase and (b) is big enough so that we have enough statistics. These parameters depend on your business. In my example I will select customers which first bought between 13 and 24 months ago.

Step 2: Choose average KPIs for that cohort


Here are the metrics I recommend:

  • Metric 1 - How many customers are in your cohort: M1=1.11M 
  • Metric 2 - how many customers in your cohort made a second -or more- order: M2=557K 
  • Metric 3 - What is the average LTV of customers in your cohort that made a second -or more- order, excluding the first order value: M3=$427

So our benchmarks for customer segments, based on first order, that indicate high value will be:

  • Average LTV excluding 1st order: (M3) > $427
  • Conversion rate to 2nd order: (M1/M2) > 50%

Step 3: Split by 1st order attributes


Take this cohort and split it by attributes you collect about your first order. We suggest:

  • First order product category (e.g. shirts vs. dresses)
  • First order sales channel (e.g. web vs. retail store)
  • First order revenue (bucketized into ~10 buckets with similar customer counts)

*Note: this can split your cohort into hundreds of segments - that's OK, we want to get granular, but that's why we need to start with a large enough cohort.

Step 4: Choose average KPIs for each segment

You will repeat the exact same calculations in Step 2 for each one of the segment (denoted here : 'i')

  • Metric 1 - How many customers are in each segment: M1(i)
  • Metric 2 - how many customers in each segment made a second (or more) order: M2(i)
  • Metric 3 - What is the average LTV of customers in each segment that made a second (or more) order, excluding the first order value: M3(i)

Select the segments that meet both of these conditions (filter out the rest):

  • Average LTV excluding 1st order: (M3(i)) > $427
  • Conversion rate to 2nd order: (M1(i)/M2(i)) > 50%

Sort the surviving segments by M3(i) descending - so the highest incremental LTV is at the top.


*This step might sound a bit gnarly, so at the bottom of this article we’ll walk through an example.

Step 5: Build the campaigns

You now have a list of first order attributes that allow you to identify potential high value customers immediately after they made their first order! You will now build campaigns to attract new customers who exhibit these specific 1st order attributes in order to get them to purchase a second time.

 

Need a SaaS solution that can help you execute strategies like this?  Request a personal demo of AgilOne Today!

Request a Demo CTA.png

 *Concrete example of Step 4

In the table below we show 6 hypothetical segments based on first order product category, sales channel and revenue, along with the accompanying metrics we outlined in step 4.

The first 3 segments are your best candidates because they exceed the benchmark incremental LTV and conversion rate from step 2, and also have high enough volume of customers (exceeding the threshold of 1,000 we set).

The best of these is segment 2 that has the highest incremental LTV and customer volume  -  start with this segment for your pilot campaign.

The remaining 3 segments each under-performs in a different metric (highlighted in green), so these are not good candidates to spend a lot of effort on - although you might argue that with a good enough incentive you might be able to increase the native conversion of segment #4 - an interesting experiment to run in your free time : )

Segment ID

First Order Product Category

First Order Sales Channel

First Order Revenue

Customers in segment M1(i)

Conversion Rate M1(i)/M2(i)

LTV excluding 1st order M3(i)

1

Skirts

Online

$50-$75

5,500

55%

$650

2

Tops

Store

$50-$75

20,224

57%

$715

3

Suits

Store

$150-$200

13,890

58%

$610

4

Bags

Online

$150-$200

3,000

33%

$560

5

Hats

Online

$25-$50

850

56%

$620

6

Socks

Online

$10-$25

12,650

75%

$128


*More refinement steps:


  • If your campaigns will be specific to each segment, that's a lot of work and you will not want to do this if the segment is very small, so you might apply a threshold on M1(i) (for example only if the segment exceeds 1,000 people will you build a campaign for it).
  • How soon after the first purchase should you start courting your new customers? The blunt (but perfectly acceptable) approach is to have your analyst calculate the median time between 1st and 2nd orders for the repeat buyers in the original cohort and use that as the time to trigger your campaign (e.g. 45 days after first order provided that the customer has not yet made a second order). But if you are an insane details-person (come on, own it!) then have your analyst calculate this median time for each segment - it is possible that people who first buy a jacket take 90 days to return vs. 30 days for a tie - you will then trigger the campaigns for the different segments at different time interval from first order.
  • Caution: if you are using first order product category as a way to segment your cohort, remember that a single customer can belong to more than one segment if he/she purchased from multiple categories. Make sure to put in the appropriate suppression rules on your campaigns to prevent one customer from being barraged by similar confusing messages.
  • What content should you present in your campaigns? It's great if you can produce content that is relevant to the first purchase the customer has made (such as complementary items for the products purchased, or new arrivals in the same category). In terms of the tone of your campaign: remember that you are targeting potential VIPs, so you might want to offer some VIP treatment: a small gift or access to private sales or consultation with a style expert.

 

Receive Blog Alerts