How to Analyse and Monitor the Health of Your Customer Base
1. How to Analyse and
Monitor the Health of
Your Customer Base
@CarmenMardiros
2. This is how it all started….
What is a “cohort”?
first_order_month customers
Jan 2016 7325
Feb 2016 2344
March 2016 2355
…. ….
3. My first cohort analysis at navabi...
My two CEOs were not impressed: “The data is wrong”
4. Aha moment - I hadn’t “equalised” customers
2013 customers had 3 extra years of lifetime vs 2016 customers!
ANY metric will look better for older customers.
5. A fairer comparison painting a truer picture.
Cohort performance is capped at X weeks of age
An emerging
trend?
Should we be
worried?
How cohorts are doing at 8 weeks of age
6. A truly fair comparison that reflects reality.
Only cohorts at least X weeks old are eligible
How cohorts are doing at 8 weeks of age
7. What age is best to measure cohort behaviour at?
Long term view sacrifices visibility into recent cohorts
but gives better insight into lifetime behaviour.
On-the-fly age capping is best depending on the purpose.
March
Cohort
July
Cohort
9. What metrics best reflect cohort health?
Leading indicators of LTV
Orders per Customer (or other actions)
Discount Rate (promos, sales etc)
Revenue per Customer
Return/Cancellation Rate
Marketing Cost per Customer (acquisition and retention)
Profit per Customer (actual and forecast)
Always capped at age X. Never measured for cohort as a whole.
Other leading indicators: website visits, products added to basket/wishlist, newsletters clicked on,
customer service tickets/complaints
10. “Per customer” metrics can be very misleading
Cohorts are rarely normally
distributed and the average can
be misleading. So:
Use median to calculate “per
customer” instead of mean, if
feasible.
Measure milestones reached.
11. What metrics best reflect cohort health?
Milestones reached
% Cohort reached 2nd order
% Cohort migrated from 2nd to 3rd order (3rd to 4th etc)
Digital equivalents: % User Cohort added product to basket within 3/7 days etc
% Cohort with discount rate over y% / avg price paid over €z
% Cohort reached profitability
(measures to what extent profitable customers subsidise the rest of the cohort!)
% Cohort reached VIP status (high LTV value)
Always capped at age X. Cross-reference with acquisition channel
for invaluable nuggets of insight.
12. Key takeaway #2
Use “milestones reached” as health
checkpoints for a wide range of cohort
behaviours.
13. Combine “calendar time” with “cohort time”
Activity in week X comes from customers at different ages. Understanding
“what works for whom” is challenging without analysing the mix.
14. Response is relative to Active Customer Base
# Active Customer base = Anyone active in previous 57 weeks
% Reorder Rate = Anyone active in previous 57 weeks and active again this week
15. Why Active Customer Base and not Total?
Removes the effect of
churned customers
Total Customer Base keeps going up and up. But it also includes an
increasingly high segment of churned customers.
A much greater re-order rate would have to happen in order to register as
strongly if we calculated Reorder Rate relative to the total customer base.
Removes the effect of
new customers
Total Customer Base usually also counts customers acquired that week.
This can muddy the picture significantly.
16. Key takeaway #3
Identify the customer segments your
organisation most depends on and monitor
their size and response like a hawk.