Recurring revenue analysis is a toolset of analysis that can be employed when a business generates revenue through subscription/contract based means.
1 Conceptual ARR guidance
2 Engagement best practices
3 Application of data analytics
4 Example analyses
2. Page 2
The significance of recurring revenue analysis
Recurring revenue analysis is a toolset of analysis that can be
employed when a business generates revenue through
subscription/contract based means.
âș Industries that lend themselves to this include software, online
services, utilities, and any consumer services that result in
memberships or subscriptions (gyms, clubs). There is a continued
shift in the software industry to move away from one-time license
sales to more recurring models of revenue making this analysis of
particular importance.
âș Having revenues earned through a recurring manner opens up a
library of analysis to tightly describe the quality of recurring
revenue and the key business drivers including:
âș Churn analysis
âș Revenue bridging (Upsell, downsell, cross-sell, new/lost
products and customers)
âș Cohort analysis
âș Customer lifetime value
âș Contractual revenue run-out analysis
âș Understanding the particularities of these analysis is critical for
building/challenging an equity story for a business with recurring
revenue.
ARR retention bridge
3. Page 3 EY TAS for Evonik
1 Conceptual ARR guidance
2 Engagement best practices
3 Application of data analytics
4 Example analyses
4. Page 4
Revenue analysis: all combinations you can think of
are both possible and relevantâŠ
Bookings / billings: TCV
& ACV
Recognized revenue
(what GAAP?)
Cash inflows
Monetary units
# of contracts
# of customers
# other project specific
SKU
SKU
Perpetual license
Term license
SaaS
Revenue stream
Maintenance
Consulting / Service
Show it how?Analyze what?
Analyze by?
Price increase /
decrease
Upsell / downsell
Cross-sell
Main views on drivers
Bridge it by?
Volume changes
New client / contract
additions
[Project specific]
Regions / countries
[Project specific]
Products / product groups
[Project specific]
Customers / customer
groups
[Project specific]
Anything elseâŠ
Direct
Distribution channels
Reseller
[e.g. Cloud, anything
else]
A) By fiscal year
Recognized in period
ARR : Contracted / recurring
amounts at cut-off date
Two perspectives:
B) Year-on-year change
C) Customer cohorts
Churn / renewal (rates)
5. Page 5
Subscription revenue analysis
Covered timeframe: Last twelve months versus other periods
ARR / MRR
in âŹ/USD
(optional: #
of SKU, e.g.
contracts,
customers)
BOP
âș Annual / monthly
recurring revenue
at beginning of
period
âș Usually value of
inventory snapshot
at a certain point of
time (e.g.31
December 2017),
can also be
recognized
âș Contracted
(âCARR / CMRR)
versus implicit
1
Upsell /
Downsell
in âŹ/USD
âș Incremental
ARR / MRR
from
increasing
utilization /
volumes with
existing
products at
constant BOP
price levels
âș Data may not
be retrievable
in all
transactions
in order to
isolate this
impact from
price
variations and
cross-selling
Price
Increase /
Decrease
in âŹ/ USD
âș Incremental
ARR / MRR
from variation
of prices at
existing
products at
constant BOP
volume levels
âș Data may not
be retrievable
in all
transactions
in order to
isolate this
impact from
upsell / down
sell and
cross-selling
Cross-
selling in
âŹ/ USD
âș Incremental
ARR / MRR
from adding
existing
products to
existing
customers at
constant BOP
prices
âș Data may not
be retrievable
in all
transactions
in order to
isolate this
impact from
upsell / down
sell and price
variations
Churn in
âŹ/ USD
or #
âș Incremental
ARR / MRR
customers
terminating
their contracts
âș Churn rates in
relation to
BOP figures
or LTM
averages may
also be
assessed
âș Definitions of
churn need to
be
understood,
challenged
and presented
Addit-
ions in
âŹ/ USD
or #
âș Incremental
ARR / MRR
customers
terminating
their contracts
âș Churn rates in
relation to
BOP figures
or LTM
averages may
also be
assessed
âș Definitions of
churn need to
be
understood,
challenged
and presented
ARR / MRR
in âŹ/USD
(optional: #
of SKU, e.g.
contracts,
customers)
EOP
âș Annual / monthly
recurring revenue
at end beginning of
period
âș It may be
necessary to put in
a mix effect as well
depending on
underlying sources
of data
72 3 4 5 6
Net retained revenue (NRR)
6. Page 6
Identifying recurring revenue streams
29 October 2020
Typical view on recurring revenue components
âș License revenue is typically non-recurring in its nature.
âș Term licenses may be a special case if they are subject to clauses foreseeing a renewal in
case not terminated.
âș SaaS (or other subscription revenue) and Maintenance is typically formed in self-renewing
(unless terminated) contracts formed for an at least annual duration. Of course exceptions
may apply and need to be challenged / raised in expert calls / Q&A with the Target.
âș Consulting and other revenue streams may also comprise certain recurring elements,
e.g. fixed hour contingents or extended maintenance fees.
âș Even if such recurring components of service or consulting projects exist, you should
raise this with the client as any man-hour based revenue is usually deemed less
valuable than SaaS or maintenance. You should decide whether to include any such
recurring other services together with the client and keep proper track of any such
disclosure decisions.
âș Especially for smaller targets, the revenue streams may not be properly maintained in
the base data. Our clients usually emphasize that any hour-based business or any other
components of service business should not be disclosed as SaaS or maintenance. You
should try to review the contents of the disclosed revenue streams for validity based on
invoice descriptions and expert interviews and adjust whenever possible / feasible.
Dimensions of recurrence
âș Recurring revenue is usually defined based on an underlying contractual arrangement.
However, in some cases, e.g. when assessing current trading or a business plan, you may
need to decide on the quality of recurrence.
âș Some SaaS contracts have a baseline revenue and a variable component dependent on
license points or other volume-indicators. For example, a contract might comprise a base
fee, a variable price per license point and a committed minimum volume growth per year.
âș Projected ARR not underpinned by existing contracts represents managements sales
ambitions.
Perpetual license
Term license
SaaS
Revenue stream
Maintenance
Consulting / Service
Other revenue streams
Recurring
revenue
streams
May comprise
recurring
elements
Quality of recurrence
Contractual
Non-contractual /
management ambition
Baseline
Volume upside
Committed upside
7. Page 7
Decision tree for establishing ARR based on
available data
29 October 2020
Minimum data requirements
âș Proper calendarization of the data (exact date)
âș Granularity of the dataset on customer level in order to be
able to assess churn
âș Revenue stream should to be traceable to judge on
recurrence of revenue (if not, billing cycles may be used as
proxy but may be misleading)
ARR available?
TCV / ACV
Billings / payment
data
Contract duration
Billing cycle data
Option 1: use TCV / ACV data
Option 3: use recognized revenue data
Cool, youâre fine ;-)
yesno
Option 2: use billings / payment data
Revenue recognized
Accrual basis
corections
What to do when ARR is not reported by the Target?
Especially for smaller Targets (e.g. owner-led firms), regular KPI reporting is often limited in
scope and depths so ARR wonât be reported at all or not in the granularity necessary to fulfil
our scope of work. There are several ways of working around this issue. Below is a brief
summary of how to deal with these issues sorted by preference and practicality:
Option 1: use TCV / ACV data
âș By dividing TCV with contract length and spreading it over the respective timeframe of the
contract equally, ARR can be computed and assigned to the respective periods.
âș ACV could also be used if maintained by the company. Be mindful of any impacts of price
escalator clauses or variable contract components.
Option 2: use billings / payment data
âș Raw billing data can also be used to calculated recurring revenue in case the underlying
contract terms can be matched to the billing data
âș In order for this method to be even more precise ,it is best to differentiate between delivery
date and invoice data. Usually the delivery date should be the first point in time for which
recurring revenue is recognized and should overrule the invoiced date or even less
relevant, the payment date.
Option 3: use recognized revenue data
âș In theory, if revenue is properly accrued on a monthly basis, multiplying the respective
monthly revenue * 12 would yield the correct ARR balance (assuming all revenue is
accrued at 1st of the month).
âș In practice, this approach has many pitfalls: Sometimes discounts and rebates are not
properly accrued on monthly basis and need to be corrected for such analysis. Also, any
true-ups (e.g. from license audits) or any other items not accrued proportionally will impact
the results.
âș It is therefore crucial when using this method to ascertain revenue recognition policies for
all relevant ARR streams / products and correct for any identifiable biases arising from it.
8. Page 8
Variants of unit economics
29 October 2020
Typical unit economics encountered in software transactions
Unit economics for software deals typically focus on contracts and customers, but other units
may also be of relevance::
âș Contracts: Base case for unit economics when TCV / ACV reporting is established.
Calculation of ARPU would usually yield meaningful results. Unit churn would also yield
meaningful results, however it should be noted that churn is more commonly defined on
total customer level.
âș Customers: Unique customers are commonly used as basis for unit economics.
Customer groups and / or families may frequently be a variant encountered in various
engagements. ARPU and churn rate computation yields meaningful results.
âș License counts: Some companies wonât be willing or able to report on licenses sold, but
in case so, they would provide for meaningful units for computing churn rates and ARPU.
âș License points: In some instances, a contract might imply a firm-wide license at flexible
pricing per license points. One example of license points would be end user access points
for a service provider where the service provider purchased the license. Usually, such
license points might not be willingly shared or readily available in a DD-context.
âș End users: May be relevant for certain commercial assessments but typically wonât be
inducing variable payments in the sense of license points above. One example of end
users might be a service tool software part of a service bundle resold by a company to
enterprise clients that also give access to the software to various users per instance. May
be important to evaluate the commercial outreach of a software but usually cannot be
reliably reported in a DD-context.
âș Invoices: Will be available when using billing data as source data for ARR analysis but
typically wonât produce meaningful result when transposing to ARPU or unit churn.
âș Orders: Will usually be available when TCV data is available. Can provide meaningful
hints on selling cycles and seasonality. Average order value can be meaningful for
analysis, order churn rates typically would not be conclusive for analysis.
# of contracts
# of customers
# other project specific
SKU
SKU
# of Licenses
# of Invoices
# of End users
# of License points
Meaningfulness of unit economics (ARPU)
âș Not all unit data will be suitable for drawing conclusions from
the figure itself (e.g. Average revenue by invoice may be
biased when comparing monthly billing clients with annual
billing clients).
âș Sometimes, such shortcomings can be neglected when
computing price / volume mix effects in case the base units
and their cyclicity do not change
A) Enables
computation
and analysis
of Average
revenue per
Unit (ARPU)
B) Enables
computation
and analysis
of unit churn
# of orders
9. Page 9
Churn rates
29 October 2020
Churn rates â computation and interpretation
Churn rates are typically discussed either on unit / logo basis or ⏠basis
Standard ways of computing churn rates:
1. Basic variant: Churn (either in ⏠or #) divided by (either ⏠or #) at BOP
2. Alternative variant: Churn / (Ă(BOP + EOP))
Defining the level of churn:
âș Usually, the identification of churn will need to be based on a specific level, e.g. contract versus
customer versus a group of customers. Generally speaking, churn rates and churn will be much
higher when assessed on individual contract level versus customer level. In case a customer has
multiple contracts, contract churn would be reflected in downsell or volume change instead of churn
in this logic. Churn and churn rates would thus c.p. always get lower the more aggregated the basis
of defining churn is.
âș It is pretty common to evaluate churn either on a customer or customer group level.
âș When establishing customer groups, checking for renaming of companies, mergers and name
duplicate is essential to validate that all customers are properly mapped to a group
Monthly versus annual churn rates
âș Churn rates can theoretically be compiled for any timeframe.
âș We noted two / three dominant forms of translating between annual and monthly churn rates. Note
that result may vary greatly depending on the seasonality of churn and new business.
âș Variant A (annual churn): Annual churn / BOP (January), alternatively / by avg. BOP
âș Variant B (average monthly churn): Sum of monthly churn / Sum of BOP per month
Assessing churn when no contract termination is recorded
Typically, when termination dates are not readily available, churn could be estimated based on the
recurrence of revenue. Any customer disappearing for more than one billing cycle would typically be
marked as churn. However, a return of the customer (âBoomeraingâ) should be checked and override
any such churn flagging.
of contracts
of customers
# other project
specific SKU
Churn
customer groups
End users
License points
#
ARPU
⏠/ $
Churning units Churn value
Variations in compiling churn rates
âș Keep in mind that there is no uniform way of
computing churn rates and so especially on the
sell-side we should understand how any churn
rate is computed and be able to duplicate this
âș Also on the buy-side the client mi
10. Page 10
Cohort analysis
29 October 2020
Cohort analysis
Benefits
âș Analyzing customer developments by time-based cohorts allows for a more
detailed view on the customer lifecycle and the development of customer
lifetime value
âș Especially in subscription based business, the analysis of customer cohorts
has become common practice
âș Also, this allows to analyze trends in sales endeavors and success of new
product launches
âș Upselling and churn rates can also be analyzed on a more granular level
Restrictions
âș Cohorts need to contain a minimum number of unit to become meaningful for
analysis.
âș Meaningful analysis is usually only possible after a certain minimum age of
the cohort
âș Ageing cohorts may be subject to certain biases in case only inactive
customers remain (which may be a finding in itself).
âș With cohorts, it is usually better to have the longest history of data available in
order to be able to form more meaningful cohorts
Typical cohort criteria
âș Calendarization: this is the typical form. Depending on
0.0
5.0
10.0
15.0
20.0
25.0
30.0
Initial
Subs.Y1
Subs.Y2
Subs.Y3
Subs.Y4
Subs.Y5
Initial
Subs.Y1
Subs.Y2
Subs.Y3
Subs.Y4
Initial
Subs.Y1
Subs.Y2
Subs.Y3
Initial
Subs.Y1
Subs.Y2
Initial
Subs.Y1
Initial
FY11 FY12 FY13 FY14 FY15 FY16
âŹm
License Maintenance & Subscription Service & Other
7.7
9.8
10.9
12.5
13.3
14.0
1.6
3.2
5.1
6.9
8.8
10.3
1.6
2.8 3.2 3.5 3.9 4.1
0.0
2.0
4.0
6.0
8.0
10.0
12.0
14.0
16.0
Initial Subs. Y1 Subs. Y2 Subs. Y3 Subs. Y4 Subs. Y5
âŹm
License accumulated Maintenance & Subscription accumulated Service & Other accumulated
11. Page 11
Customer lifetime value
29 October 2020
Customer lifetime value
Benefits
âș Customer lifetime value (CLTV) and its comparison to customer acquisition
costs (CAC) is a key metric for recurring revenue business that is used by
buyers and financers to appraise the health of the business.
Restrictions and considerations
âș Customer lifetime (CLT) is typically estimated by taking the reciprocal of the
average churn rate. E.g. average annual churn rate of 33% results in a
customer lifetime of 3 years (1/33%). Cohorts must be compared over similar
terms to arrive at their average churn rate, particularly when churn rates
decrease rapidly year on year.
âș Customer Value (CV) should be provided from the lowest subdivision of
commercially meaningful customer available. E.g. one estimated monthly
value for all customers on average wouldnât be very helpful. Having a value
for each cohort and contract length would allow for the identification of
particular trends.
âș Customer acquisition costs (CAC) are costs directly attributable to acquiring
new customers, this is usually al costs associated with marketing to
customers, paying rebates to brokers and other incentives to third parties in
the customer pipeline.
CLTV of Jan-Cohorts over time
Jan16 Jan17 Jan18 Jan19
3 month 22 20 20 19
6 month 26 25 24 22
12 month 34 32 30 31
weighted avg. Lifetime
(months)
25 25 26 25
3 month 9.5 9.4 11.1 10.9
6 month 7.8 7.7 8.4 8.0
12 month 5.3 5.3 5.8 5.2
weighted avg. monthly
value (âŹ)
7.6 6.6 7.1 6.8
3 month 209 184 222 203
6 month 205 190 200 172
12 month 178 168 177 162
CLTV (âŹ) 191 168 186 170
CAC (âŹ)Âč 61 43 35 28
CAC multiple 3x 4x 5x 6x
Customer
Lifetime (CLT)
Customer Value
(CV)
Customer
Lifetime Value
(CLTV)
Customer
Acquisition Cost
(CAC)
CAC Multiple
12. Page 12
Contractual revenue runout
29 October 2020
Revenue run-out by subscription duration with renewals
Ref: Gymondo BI System Download, Financial Model, EY Analysis
Currency: ⏠000 1 month 3 months 6 months 12 months Total
Jul-20 9 619 287 1,118 2,031
Aug-20 8 524 267 1,098 1,897
Sep-20 7 475 255 1,081 1,819
Oct-20 7 422 235 1,064 1,727
Nov-20 6 391 215 1,040 1,652
Dec20 6 367 202 1,025 1,601
FY20 43 2,798 1,461 6,425 10,727
10,726
21,453
23,179
2,840
1,726
7,887
FY20BYTDJun Contracted
Revenue with
renewals
Total Remainder
10,727
FY20 budget bridge after contracted revenue and renewals
Contracted revenue runout
Benefits
âș As a result of the reliability of earnings for businesses with recurring revenue,
budget analysis can be much more measured and precise. In the example
left, it was shown on a live engagement that as of June 30 2020, 93% of the
yearâs full budget could already be considered earned using contractual
revenue run out analysis.
Restrictions and considerations
âș Detailed contractual revenue run-out analysis requires specific knowledge of
each live contract and its remaining term. These remaining revenue days for
each contract can then be turned into revenue to be realized via their
contracted average revenue per day. (This is the amount presented in gold in
the bridge).
âș In addition to simple contracted revenue, the analysis can be extended to
allow for renewals of those contracts based on historical retention rates.
âș After a subscription expires, the next monthâs revenue associated with that
subscription is equal to the prior monthâs revenue scaled by the probability
that the user renewed based on historical renewal rates.
13. Page 13
Typical dimensions (excluding revenue stream) â
Region, customer, customer group, brand, channel
29 October 2020
Typical dimensions used in analysis (âthe stuff to build slicers and filters fromâ)
As a general rule, never delete dimensions from a dataset if not technically necessary. Typically, if data was disclosed,
client might request an analysis and it is always easier to omit than to append.
1. Geography
âș City
âș Country
âș Region (combination of countries, may vary from client to client, make sure to disclose definitions)
âș When presenting a split by geography, it should be understood how this dimension is determined. Typical variants
include:
âș Invoice address (which may not be reflective of actual geography of the end user)
âș Selling legal entities' address (which even more so may not be reflective of actual geography of the end user)
âș Other data sources to be explired on case-by-case basis
2. Sales channels / agents
âș Sales channels are typically split into direct sales and indirect channels such as resellers or OEMâs who embed a
targetâs products.
âș An analysis by sales agent will be highly insightful if feasible to asses key sales team members to keep / incentivize.
3. Products / product groups / brands
âș Products / product groups and brands can usually be established in a mapping table.
âș Multiple n to n relations might occur which without a mapped flat file would limit the flexibility of any analysis
3. Customers / customer groups / size / cohorts (see previous slides)
4. Invoicing currency
âș Billing data and contract data typically includes information on currency. This is highly valuable for analyzing the FX
exposure and impact of FX fluctuations on sales / EBITDA and to perform constant currency analysis
5. Other dimensions may be established on a case-by case basis
Geography
Sales channels
Customer / customer
groups / size
Typical dimensions
Sales agents
Billing cycles
Brands
Products / product
groups
Departments
Cohort
Invoicing currency
Deal size
Other dimensions
14. Page 14 EY TAS for Evonik
1 Conceptual ARR guidance
2 Engagement best practices
3 Application of data analytics
4 Example analyses
15. Page 15
Standard software DD request list
29 October 2020
Notes to standard request list
âș The below attached IRL contains all typical FDD items as well as a sales cube download request which should suffice to prepare a market standard ARR
analysis
âș It is however best practice to have a call early-on with the target in order to evaluate âthe art of the possibleâ for each item and explain aim of analysis and
reflect on potential workarounds based on actual data availability.
16. Page 16
Request lists for ARR analysis
29 October 2020
Customer ID Product ID Volume (if relevant) Contract ID Contract Start Date Contract Term
(Months)
Contract Value
Customer 1234 Product 9876 3 Contract 9999 20/12/2019 24 1000
Core table - Option 1 (preferable)
All contracts, per customer, per product/service, showing contract start date, contract length and total contract value.
Core table - Option 2 (if total contract value is not available)
Detailed transaction level data by customer, product/service. Data should be transactions dated to the day. We will use this data to arrive back at a table
that looks like the preferred table in option 1, so any contract information available to help minimize assumptions is desirable. E.g. the same customer,
paying the same amount, for the same product every 3 months will usually be assumed to be on recurring 3 month contracts for that product.
Customer ID Product ID Volume (if relevant) Transaction date Transaction Amount
Customer 1234 Product 9876 3 20/12/2019 1000
Supplementary tables (required in all cases)
a. Customer IDs with customer acquisition dates (we would expect these dates to exceed the scope of the transaction and the transaction
data provided for many customers)
b. Roll-ups of customers into any business relevant cohorts, e.g. industry, region, channel, size (SME, individual etc).
c. Roll-ups of products/services into business specific product/service groups
Optional extras:
If teams consider constant currency to be important then the core tables should include the transaction currency
If teams consider the legal entity selling the product significant to the analysis then this field should be added to the core table request.
17. Page 17 EY TAS for Evonik
1 Conceptual ARR guidance
2 Engagement best practices
3 Application of data analytics
4 Example analyses
18. Page 18
Workflows improve standardization,
automatization and harmonization
29 October 2020
Analytics tools based on workflows such as Alteryx,
Power Query or Python support standardization and
automatization of recurring revenue analyses.
Speeding up the analysis, the data process workflows
ensure a consistent deliverable across projects by
pre-configuration and conformity to consistent
definitions.
Sales-cube data received from the client for the
purpose of financial due diligence are often in a
similar shape and structure with data extracts only
needing little adjustments or amendments before they
can be fed into the workflow. So standardization
actually works!
Workflows are useful for data cleansing, manipulation
and preparation. The resulting flat files are the basis
for customized or pre-designed MS PowerBI reports
with several pages of analysis.
19. Page 19
Dynamic dashboards slice analyses flexibly and
allow further granularity
29 October 2020
A breakout of the ARR
categories in a matrix
format.
Evaluates New in
relation to Lost, on an
ARR $ and logo basis.
Evaluates total
ARR $ per active
customer.
Evaluates New
ARR $ per New
customer.
Evaluates Lost
ARR $ per Lost
customer.
Displays $ and
logo churn over
time. When a
selection is made
(e.g., Cohort),
the âTotalâ
ignores filters to
show the rates
for your selection
in relation to the
Company totals.
Displays ARR $
and NRR% over
time. When a
selection is made
(e.g., Cohort),
the âTotalâ
ignores filters to
show the rates
for your selection
in relation to the
Company totals.
Values in the below
visuals should be filtered
to the last date in the
dataset. The NRR%
gauge shows the Dec18
NRR% relative to the
Min and Max for the
Historical Period.
Rapid preparation of recurring revenue analysis topics including churn and revenue bridging can be achieved in a standardized way with data models and dynamic
dashboards such as PowerBI or Excel Power Query. Slicers allow to quickly focus the analysis on a special topic of interest and view the data from a particular viewpoint.
20. Page 20 EY TAS for Evonik
1 Conceptual ARR guidance
2 Engagement best practices
3 Application of data analytics
4 Example analyses
23. Page 23
Report visualization of ARR bridge
29 October 2020
Client A (âŹ736k)
Client B (âŹ591k)
Client C (âŹ371k)
Other Top20 (âŹ359k)
Remaining Upsell (âŹ832k)
Client A (âŹ900k)
Client B (âŹ530k)
Client C (âŹ459k)
Other Top20 (âŹ383k)
Remaining Upsell (âŹ1,252k)
Client A (âŹ534k)
Client B (âŹ389k)
Client C (âŹ258k)
Other Top20 (âŹ437k)
Remaining Upsell (âŹ1,357k)
Client A (âŹ1,055k)
Client B (âŹ297k)
Client C (âŹ224k)
Client D (âŹ196k)
Client E (âŹ181k)
Client F (âŹ115k)
Other Top20 (âŹ318k)
Remaining Upsell (âŹ928k)
Upsell Downsell New Logo