1. Cassie Lancellotti-Young
VP Client Analytics, Sailthru
May 2013
@dukecass
Metrics that Matter:
The 360-Degree Customer
Transcending Data Silos for Holistic Marketing
2. About Me: The “Reader‟s Digest” Version
media/tech banker at Citigroup
acquisition and subscription analytics at TheLadders
independent analytics consultant while MBA‟ing
marketing and analytics at Savored (exited to Groupon)
intrapreneurship at Gerson Lehrman Group (GLG)
2013
2005
client optimization/analytics at Sailthru
8. the ugly:
the ignorant marketer
“Cassie, check out our updated mobile app”
Cassie‟s account is
already linked to an
iPhone app
No options to upgrade
AMEX app in App Store,
so Cassie likely already
has this version
9. 39% of marketers
can‟t turn data into action
Stat Source: Columbia Business School: “Marketing ROI in the Era of Big Data,” 2012.
Image Source: Dilbert, 29 July 2012.
20. give yourself a sanity check!
(b2b vs. b2c, price point, etc.)
21. measurement happens via 3 lenses
user level – what are users doing?
product/transaction level – when is site
conversion highest? which products yield
the strongest repeat rates?
relationships/ratios – how do certain
experiences impact user behaviors (e.g.
first purchase type vs. NPS)?
23. always have a pulse on
what happens on downstream
Source: Monetate
24. just say no to “elevator” analysis
understand the why?
product/marketing: deliberate changes to
messaging, site, etc.
business ecosystem: i.e. inventory issues,
technical problems
“macro” factors: i.e. industry trends,
economic climate, press
26. live chat usage vs. conversion
TheLadders saw a fairly immediate 13%
increase in premium conversion for a test group
that live-chatted with a representative.
27. feedback vs. subsequent use cases
Some Savored restaurants were absolutely
horrid for driving repeat usage.
28. account management vs. spend
Does more time spent on account management
yield upsells? Higher contract value?
Does it reduce churn?
29. tying together online/offline behaviors
Customers who try the brand offline will
oftentimes have higher AOVs (average order
value) online downstream.
30. “magic” numbers
define tipping point for engagement
Facebook considers a user to be “engaged” if
s/he gets 7 friends within 10 days of signup.
31. understanding relationships is as easy as 0,1
(0=no, 1=yes)
Data science/predictive
models are most ideal, but
you can get started
(directionally) on your own
with simple binomial
regressions.
e.g. Savored regressed
restaurant churn propensity
against reservations in first
30 days.
45. so your customers don‟t pay you?
not a problem…
Develop proxies for revenue – a post, a
Tweet – but make sure those proxies
are truly valuable behaviors.
For advertising-driven businesses, still think
about key value – i.e. PV/user yields $X in
ad revenue?
60. old school segmentation has evolved
(because customers aren‟t segments)
MANY PEOPLE
Recency
Frequency
Monetary Value
“80% of your revenue
comes from 20% of
your customers”
CASSIE‟S MODEL
Behavioral
Usage
Situational
“what else do we
know about that 20%
segment?”
69. big data vs. big brother
“your friend Amanda dined at Zengo with Savored”
Disengaged Segment
10% open rate
15% CTR
Dormant Segment
30% open rate
10% CTR
73. again, the right metrics are
often more of an art than a science
Twitter‟s success
metric for this email is
likely something like
“incremental follows”
74. make it easy: which ONE thing
do you want the user to do?
?!?!?!?
75. this is pretty straightforward
(probably optimized for day1 buyers?)
76. don„t forget about
creative optimization
Savored – increase from 4-15 restaurants
per email increased RPM by over 300%
ASK ROOM who is from startups vs. established brandsbusinessanalytics are a science, but devising which metrics are mission-critical for your business is really something of an art
38% of US adults are always addressable2/3+ of people under 45 are always addressable
industry definition -4Vs = volume, velocity, variety and veracity
Improving the signal-to-noise ratio
Lets you look at things like – X% lift in Facebook followers last Tuesday, but did that also result in increase to subscribers? Near-term purchases?
And of course, there are myriad tools that will plug into database and let you slice/dice on your own – Tableu, ProClarity, etc.Think of these as pivot tables on steroids
User lookup – basically a B2C CRM tool to see everything about the customer in one place – devices, whatever custom vars the client was to track, etc.
1/0 approach is great for synthesizing data via pivot tables – we’ll talk more about 0/1 scoring later
Someone can feasibly use OpenTable multiple times per week, the same cannot be said for airplane ticketsDay1 purchase behavior – if that were relevant in B2B, many sales/AMs wouldn’t have jobs
everyone should (hopefully) be looking at the first 2, but analytics superstars are hyper-focused on the thirdratios/relationships are really the crux of non-siloedbusiness analytics
Vanity metric would be aggregate customer conversion rateDifferent email treatments can be a cohort of users – how does the behavior of someone who receives welcome series A differ from someone who gets Series B?Relates to product level as well – how does someone who gets a product tour differ from someone who does not?Can also use on the B2B side and look at churn rates…
Not always about immediate results – think about what happens to someone who gets a welcome series with a discount – might activate sooner, but downstream LTV might be lower – in this example, we don’t care if people only buy with free shipping if the amount they always spend makes up for it profit lost
Elevator analysis – “it went up” or “it went down” – might be interesting to hear, but completely uselessReiterates importance of ratios/relationships – explains the WHY behind those metrics…
In case you still have no clue what I am talking about…Going to highlight a # of examples – these are by no means comprehensive, but will help you think about what to look at
In early days of live chat at TLC we saw a 13% increase in conversion from basic to premium product for group that live chatted with a representative
Post-dining survey to users after each experience asking them to rate experience at restaurantWe looked at feedback scores from first-time users and then mapped that to repeat rate and learned that some restaurants were more likely to inspire better customers, so only featured these restos in our early marketing materials
Do more calls/higher time spent mean more $$?
Revtrax – ties online coupons to online spendingIdentify cohorts of users who came in during the time a TV spot aired
Facebook – 7 friends within 10 days of signing upZynga – D1 retention – if someone comes back a day after signing up, good indicator… SUPER IMPORTANT because “1 and doners” are big problem
0/1 – for every X increase in reservations, risk of churn was reduced by Y%Savored example – likelihood of churn based on # reservations in first 30 days. MULTICOLLINEARITYIf there is one thing to know about stats, its SIGNIFICANCE TESTINGI’m not going to stand here and talk about actionable ordered logistics though, want to jump into quicker more actionable metricsHOW WOULD YOU GET STARTED, people ask. The answer is Dave McClure!
There are so many possible metrics we can look at, how we do think about approaching business analytics in an organized way?Going to borrow a page from Dave McClure.
Different channels will interact differently with landing page examples – branded search terms vs. fleeting message in display adsIt’s helpful to isolate these metrics (e.g. compare CTR for creative), but it’s also important to look at what is happening downstream – look at cash/keyword ROI, etc.
Are they not signing up because the form is so long?WHAT THIS ALL BOILS DOWN TO…
Understand the difference of different landing pages – reg/chall example from TheLadders with no paid option on first stepTHIS IS NOT NECESSARILY A PURCHASE CONVERSION PROBLEM!
Think about difference between B2B and B2C!
some companies (especially B2B) will even take this a step further:what is the TOTALcost of retaining a customer (account management, etc.)?COMPARE THIS # TO YOUR LTV to understand the unit economics of a customer
This is important because app owners may have higher LTV, etc.People most likely aren’t going to submit a lead gen form for a bakery, so click-to-call is a nice way to measure efficacy of spend
Downstream metrics – CPA might be higher, but LTV may also be much higher
Facebook example – can’t say the channel isn’t valid, might just be that the East Village ad eating 80% of your budget does not perform but the AARP-targeted ad at 5% allocation is crushing it
focus on mechanics of the lower funnel before pumping the gas on acquisition – squeeze every nickel until it screams
i.e. forcing users to follow things is not a valuable behavior
Same thing goes for content sites – how much is available/depth of content when someone first comes to site?
Beyond supply and demand, think about the staging of your productBirchboxor any “waitlist” product – would beinteresting to regress wait time vs. lifetime purchase behavior
Discount offers for cart abandonment or product re-targetingLots of people doing retargeting – Facebook, etc. Can also do push notifications, etc.
We called people at Savored to understand why they weren’t booking, then scored those users to understand downstream behaviorTLC – welcome call to premium subscribers, lift on renewals
Important to understand the trade-offs between price promotion and long-term value – let’s think about a sample subscription site
Accomplished via email/mobile/ads (push), product marketing (pull) and developing positive customer satisfaction
Historically marketing was about 1:many and 1:1, where the latter mostly dealt meant transactional emails or cold callsThis is why there has been an explosion of companies focused on persuasion architecture or personal experiences – monetate, etc. that were focused on 1:FEW.NOW there are tools to let us customize our content with low development resourcesPeople focus on a lot of types of targeting – geographic (so can put in things about the weather), etc. but behavioral tends to render the biggest bang for the buck
Seamless used Device Targeter to double CTR on app download prompts and to increase app downloads by 50%....And because we love DOWNSTREAM metrics, they also saw that clickers placed 90% more orders after clickers
I had opened the app, but not redeemed any deals
This is a model I made up, so you don’t have to listen to it, but…RFM analyses say nothing about what is driving customers to make purchases… so many precursors – how often are they being emailed, do they have app, etc.
A lot of people will ask me what are some easy retention wins? I want to stay focused on metrics but will quickly power through a few examples that leverage email but keep in mind you can also leverage what we talked about earlier – remarketing across the web, etc.
But make sure you are taking all of their behaviors into account – i.e. what if they haven’t been on the site, but have been in the app?MEASURING efficacy – 3 groups – this segment, segment with this email and no offer, segment that gets nothing Isn’t this kind of like RFM? DON’T JUST DO IT IN BATCH SENDING – set this up as a trigger
Engagement – what else they look at/how deep is sessionLift – do people go on to buy more, etc.Beware of the filter bubble – NYT example
Warby Parker example?Particularly worth trying with people who viewed a ton of pages and bought nothing
“hurry,” “last chance,” etc. always do well
At Sailthru, the typical client sees a 6% in monthly email revenue when they use cart abandonmentConsulting for HA – drove over $7MM in incremental annual revenue via booking flow abandonment
Can create drip campaigns so you’re always coming back to see what friends are doing – Quora does this well
Seems like a reasonable place to talk about email measurement, where much of the science is pretty antiquated
So Twitter could have “follows” as a goal in GA and tracks conversion of this email
either business goal related (get revenue)Or could be engagement goal related (get dormant people to do something, etc.) – talk about Savored game
If the product isn’t converting, email won’t do anythingAlso look at source – saw 0% conversion from iPhone – there was a bug!Is the funnel as simple as it could be?
VIRAL is not a strategy because you can’t guarantee it, but it can helpIs LTV higher for customers who were referred? ENCOURAGE MORE REFERRALS – if there are different ways to spread the word, which are most effective?
Social APIs
Revenue is a function of much of the behaviors we’ve already discussed
Jason Goldberg’s blog – 10% purchase conversion in first week from iPad users – and they are forecasted to be worth 2x more over the lifetime
If I were a betting woman – app users use more often after last purchase, so how to drive more usage through thatCAVEAT:Understand what is truly causality (for instance, do people spend more because they have an iPad, or are they already power users so will just leverage any vehicle possible to buy more?)