SX Live Events
Max Greene 25 min

How CSMs can benefit from predictive support signals


In this longer edition of Ask Max, Max uses his experience as a CSM to show all the ways the role can benefit from predictive support signals and SupportLogic workflows.



0:00

What kind of value can customer success get from the signals that are seen in

0:05

support?

0:05

And this is obviously given that support talks to your customers more than

0:09

anyone else at your organization.

0:11

As a result, there's a wealth of information in those interactions.

0:16

And the worst thing that can happen to a CSM is to walk into a conversation

0:23

with a customer

0:24

completely blindsided about the fact that the world is already on fire and not

0:30

knowing about it.

0:31

And depending on how closely your CSM teams work with your support teams, this

0:36

can absolutely happen.

0:38

At SupportLogic, obviously, given that this is our whole value proposition,

0:44

I work very, very closely with my support team.

0:48

However, even then, the last thing I want to do is spend all my time reading

0:52

through support tickets.

0:53

And supportLogic very much highlights the ones that are important to me for my

0:58

book of business

0:59

so that I can stay on top of it.

1:01

So that's a little bit of a high level of what we're going to talk about and

1:06

look at today.

1:07

But to give a slight bit more about myself, I've been a CSM for...

1:15

[laughs]

1:16

Right, Stephen.

1:16

Yeah, I've been a CSM for the last, gosh, about eight years now, seven or eight

1:23

years.

1:23

And I've been working in the support technology space at the different

1:29

organizations and companies

1:30

I've worked at for about 10 years.

1:33

So this stuff is very near and dear to my heart.

1:37

So with that in mind, let me...

1:41

I'll go ahead and share out my screen and talk a little bit about what we do,

1:45

what SupportLogic does and why it's so valuable.

1:50

So for those who don't know, and I think most folks have a little bit of an

1:56

idea if you had

1:57

the pleasure of coming to our booth last week, unless you were playing Space

2:01

Invaders and Dig Dug

2:02

the whole time, which was certainly an option.

2:06

SupportLogic is very much a natural language processing and ML driven, in a

2:16

sense, analyzer

2:18

of all of your support cases, right?

2:21

Reads them end to end.

2:23

We're talking every single comment, every single internal note, all the

2:31

communications

2:32

and reading that and extracting voice of the customer from it.

2:37

So distinct signals that we might detect on cases.

2:41

And as I mentioned, this is super critical for the simple reason that

2:45

Support talks to your customers more than anyone else.

2:48

So before I jump into some very specific use cases for customer success,

2:54

it's super... it's important to note that some of the types of signals that we

2:59

're looking at,

3:00

that we're categorizing across your customer conversations, range from very

3:07

procedural language

3:08

that customers might use, which are things like what you're seeing here.

3:13

And let me... maybe I can... I wonder if I can zoom a little bit more here so

3:18

we can...

3:19

Yeah, okay, there we go.

3:22

So very procedural language, something like here, you'll see this is a signal

3:27

we categorize as

3:28

critical issue.

3:29

Customers saying that they have a serious issue, maybe their system is down,

3:34

something like that, things like an escalation request when they're explicitly

3:38

asking for an

3:39

escalation, follow-up requests, they're asking for updates again and again, or

3:44

just expressing urgency.

3:45

As well as a number of others, right?

3:48

We have another one, production issue, and also call requests is a common one

3:53

that we see as well.

3:54

And the long and short of it is that some of these signals can be super

3:59

valuable even to me as a

4:01

CSM. And while I'm not going to spend all this time necessarily in this

4:05

platform going through

4:06

the cases one by one, there are situations where I want to be alerted.

4:12

And when I'm alerted, it can draw me into here.

4:14

And we'll touch on that a little bit more in a moment.

4:16

But so these are some of those procedural language type signals.

4:20

I call them procedural language because they're not emotional language.

4:23

But your customers express emotional language in support cases as well.

4:27

And so we categorize those slightly differently, right?

4:32

Baseline negative sentiment, which is kind of a catch-all for relatively

4:38

standard language

4:39

of a customer saying that they're upset or worried or things of that nature.

4:43

Frustration, right?

4:45

It's also pretty self-explanatory, but frustration actually correlates

4:50

extremely closely with

4:53

very, very poor CSAT scores.

4:55

More so than any other negative signal that we detect, if one of your customers

5:00

is getting

5:01

frustrated in their engagement with support, that will typically start us on

5:06

the path towards

5:07

an escalation, an escalation that will make its way often back to me through it

5:13

being surfaced

5:14

to say an exec sponsor that I work with at one of my customers.

5:18

It's happened on more than one occasion for me.

5:21

Though it happens less now that I'm leveraging support logic in this capacity.

5:25

So the number of signals there, right?

5:28

We look for profanity, right?

5:30

We look for impatience, confusion, super valuable.

5:34

Confusion is interesting because it's valuable not just, you know, it's

5:39

valuable in a number of ways,

5:40

right?

5:40

Perhaps your customer is confused about how to use your product in certain ways

5:48

This creates the opportunity to look at enablement for your customers, which is

5:53

definitely

5:54

an opportunity, it's a monetization opportunity potentially, or it's simply an

5:59

opportunity to

5:59

drive more adoption.

6:00

And more adoption typically leads to more growth and potentially expansion,

6:07

creates

6:07

revenue opportunities.

6:09

So these signals can be really leveraged in a number of ways to assist in

6:16

making these

6:17

decisions and understanding these things.

6:20

Okay, so that's a very kind of, that's kind of the 20,000 foot view, I would

6:26

say, of

6:27

support logic, signal extraction, and the individual signals as well that we

6:31

look for in our analysis.

6:33

So let's take it a step further to think in terms of how do I use this day to

6:38

day, right?

6:38

I don't need another UI, honestly, to live in.

6:42

I live in a lot of UIs already, right?

6:45

So the beauty of what support logic does is it allows me to set up alerts to

6:53

alert me in some

6:53

of these scenarios via my messaging channel of choice.

6:57

That support logic we use Slack for this purpose, right?

7:00

So you can be very customized in the types of alerts that you build.

7:06

There are scenarios here that I absolutely want to be notified.

7:11

So if I jump over to say our alerting page, I can configure alerts when my

7:17

specific book of business

7:19

and cases associated with those customers within my book of business are

7:26

showing that the customer

7:28

is, are showing negative sentiment from the customer or an escalation being

7:32

predicted.

7:33

I absolutely, literally need to know these things before I'm getting on a call

7:39

with one of my

7:39

customers. And so I have some of these alerts set up.

7:42

And I'll actually show you a little bit of what some of these look like within

7:48

Slack.

7:48

And this is obviously, it's not just available in Slack, it's available in MS

7:53

teams as well.

7:54

And let me, I'm going to switch my share real quickly here.

8:00

Here's a few examples of the types of alerts that I can get, right?

8:09

So here's a really interesting one. Here's a case for an, can I zoom again here

8:17

Yeah, there we go.

8:18

Right. So here's an example of a case for, let's say a customer, this Kansas

8:24

City Chiefs,

8:25

it's a football team obviously, not one of my real customers.

8:29

But you'll see here, it's telling me that a case from this customer is likely

8:34

to escalate.

8:35

And it highlights some of the key contributing factors to that, so that I know

8:40

what's going on,

8:41

right? It actually is showing me that the account health is bad. And I'll talk

8:45

a little bit more

8:46

about what that means and how support logic calculates that momentarily. But

8:51

the most important

8:52

thing to keep in mind is that when we're talking about support signals, account

8:56

health is really

8:57

the support health of a customer. Super critical, because, you know, as I

9:01

mentioned,

9:02

our support team talks to our customers more than anyone else.

9:06

We have these sentiment scores and attention scores, sentiment scored denoting

9:11

how customers

9:11

are really feeling about a case, and a neutral sentiment scores about 70. So as

9:16

it's trending

9:17

down below that, you'll see we color coded as yellow. Yellow isn't, you know,

9:20

the worst thing

9:22

just yet. But once if it continues down below about 30, it's going to get into

9:26

the red, right?

9:27

And that usually means that we've detected a number of those negative sentiment

9:31

type signals

9:32

on it. And then the needs of the attention score is based heavily on those

9:36

procedural signals that

9:38

I was talking about, but also based on like, is my team responding to this

9:41

customer?

9:42

So this is just one example of an alert, right? And you can configure these

9:50

alerts, obviously. I

9:50

don't want to be notified for every single customer of support logic, right? I

9:55

need to just be notified

9:56

for my book of business. I manage about 14 accounts here at support logic. And

10:01

so I can't be reading

10:02

all the cases for those accounts. And not only do I not have time for that, it

10:06

's not a good use

10:07

of my time. I'm a CSM, which means that my activities really need to be those

10:12

of action, right? I have

10:15

outcomes that I'm trying to make sure that customers achieve. And but on more

10:19

than one occasion,

10:20

I've worked with a customer over my career. And because of how many support

10:23

issues are open,

10:25

they don't want to talk about outcomes. They just want their problems fixed,

10:28

their problems resolved.

10:29

So, okay, so as mentioned, here's an example, right? You'll see some other

10:35

alerts here, right?

10:36

This one is specifically for a high needs attention score suggesting that this

10:40

case needs,

10:41

you know, really does need attention. I can always click for additional details

10:45

there.

10:46

But I think the key thing here, once again, showing me account health shows me

10:50

the case owner,

10:51

right? Because that can be important because if one of my key stakeholders has

10:57

gotten involved

10:58

in this case or has even opened a case, I've actually had situations where I've

11:01

seen an executive

11:02

sponsor of mine open a case, right? And I can have alerts configured

11:06

specifically when those

11:07

individuals open cases. And I've done that before so that I know. So another

11:12

great way for me to

11:13

stay apprised of what's going on in support without actually looking at the

11:17

tickets, right? I can

11:19

just see them, be highlighted, have them highlighted to me from here in the

11:22

learning capacity. And if

11:24

I really want to know more beyond that, I do have the option to click one of

11:29

these and actually go

11:30

directly into the case and support logic and see the details. I think actually,

11:36

if I were to jump

11:37

back here and this is going to take me back into support logic where I can

11:43

actually see the details

11:45

of the case. Ryan, can you confirm? Are we looking at support logic? And now

11:50

did we shift from Slack

11:50

to support logic? Yes. Yeah. You're on the Detroit Lions case. Awesome. So this

11:56

is just a little bit

11:56

of what's possible, right? But I think one of the key things we mentioned,

12:00

right, was that I mentioned

12:01

just a second ago was account health and how account health is calculated. So

12:06

we look at the

12:08

support experience as a whole and we're able to generate this support call it

12:13

account health, call

12:15

it support health score, right, which I think is a little bit more accurate.

12:19

That gives me an idea

12:20

of the customer's current experience and support. It's more than just the case

12:23

sentiment, which on

12:24

this particular customer we can see is trending down, has trended down across

12:29

their cases over the

12:31

last 30 days and it's actually pretty low. 33 is not great, right? But it also

12:35

tells me how many

12:36

escalations has my customer had opened recently, right? What's their typical

12:40

case age? And is that

12:42

going getting better? Is it getting worse? How many tickets does my customer

12:46

have opened that has

12:48

have defects tied to them? Engineering issues tied to them, because that's a

12:52

totally different

12:53

conversation, right? My customers opening a lot of like cases, asking how to

12:59

questions, that's great.

13:01

They're engaged, right? But if they're opening a lot of cases that have defects

13:05

tied to them,

13:06

that's really problematic, right? It means that maybe our engineering team is

13:12

not addressing issues

13:13

in a timely fashion, not solving for bugs, right? The last thing I want to do

13:18

is have a

13:19

conversation with my key stakeholders where they're basically telling me, you

13:22

know, your product

13:23

just doesn't work, right? That's really problematic to me. So a lot of great

13:30

insights here that roll

13:31

up into this calculation of this account health score and give me that snapshot

13:36

that I need.

13:36

Also, you know how I also use this to use this for QBRs and I hate the term QBR

13:40

, right? Support

13:42

logic we typically refer to them as outcome reviews, right? Because the CSM's

13:46

job is about

13:47

helping a customer achieve their outcomes. But I like to include as a slide in

13:51

my presentations

13:53

to customers on those reviews, I like to include what I know to be their

13:58

support experience with

14:00

support logic. I like to validate that with them, right? And being able to do

14:05

that with raw,

14:06

with like actual data, conclusive data that I can share with them can be super

14:12

valuable, right?

14:13

And it also allows me not just to, right, ideally, I'm presenting a very

14:18

positive health score

14:19

with positive results, you know, and positive outcomes that the customers are

14:24

having. But

14:26

being able to keep track of this and be alerted on it allows me to also pro

14:29

actively escalate

14:31

situations, right? Allows me to reach out to a customer proactively reach out

14:36

to one of my

14:36

exec sponsors, which I have done before based on these details, just to tell

14:41

them that I know

14:42

that there's a serious problem. I'm ensuring that all hands are on deck at

14:47

support logic,

14:48

working to address this problem. And I can often do this. And in this actual

14:56

scenario,

14:56

I let my exec sponsor know this before his team let him let him know this,

15:00

right? And you could

15:01

look at it two ways. Maybe I don't want it to. Maybe I don't want to. Maybe

15:06

maybe the situation

15:07

would have gotten resolved before he's before the exec sponsor found out about

15:11

it. But I'd

15:11

rather be proactive. Let them know that we're on top of it, that we take their

15:15

business seriously.

15:17

So this is I went on kind of a lengthy, lengthy monologue there. But I think

15:25

that gives a good

15:26

picture of what we do of what we do, not just at a high level, but at a kind of

15:30

a more detailed

15:31

level, right? Sentiments, alerting, account health, actionable information that

15:39

can be passed to a

15:40

CSM via Slack, via MS Teams, via email, via any number of other channels, so

15:49

that I can very much

15:50

take an action based on this information to ensure that my key customers are in

15:54

a good spot, right?

15:55

Because maybe a renewal is coming up. Maybe I'm working an expansion

15:58

opportunity, all these things.

16:00

Let me stop there. That was extensive.

16:05

Max, that was awesome. Thank you for that tour and sharing these things. We've

16:09

got a couple

16:09

questions in the chat here. Jordan's got two questions for you. I know Jordan.

16:16

Yeah. Jourdan's a customer of ours. That way is the customer of ours and Jordan

16:21

, nice to see you.

16:22

Absolutely. Jordan, one of the things we were showing and talking about at

16:30

Pulse last week

16:31

is our gain site integration. We do have a gain site integration that injects

16:36

certain of these

16:37

signals directly into gain site, into the timeline, so that calls to action can

16:44

be created,

16:45

right? That account health can be shown in gain site. We're absolutely have

16:52

that capability

16:53

right now. I think I chatted with your, I don't want to say your boss, but I

16:59

think your boss,

16:59

Sean, at gain site about this at the Pulse Conference last week about this

17:04

specifically,

17:04

right? And some of these other key things.

17:08

So, yeah, super valuable to your other question, Jordan, about the account

17:15

health score alerts.

17:16

Still working explicitly on those, right? Expect to have that come together

17:23

this quarter fully.

17:25

So, yeah, thanks Max. You're right. We have some flexibility with that timeline

17:30

custom object. There's different ways that these alerts, these sentiment

17:37

signals can come in a

17:38

gain site. So, that's a really cool thing that we're doing. It doesn't look

17:42

like we have any other

17:43

questions going on in the chat. Let me touch on one other thing, right? I know

17:47

that not everybody,

17:48

and even not everybody at Pulse when we were there last week actually uses gain

17:52

site, right? There's

17:53

a number of other CS platforms that companies use to tango the tally. There's a

18:00

lot, there's a

18:00

lot of them, right? I think it's super important to note, while we are working

18:06

towards direct

18:07

integrations with some of those, I think Tatango is the next one on the list.

18:12

But our CS integrations,

18:14

our gain site integration, and all these others, they're built off of our

18:17

events API,

18:17

which is our alerting API. And so, technically, right, and that's an external

18:22

facing API that can

18:24

be used to generate all of these signals that we're putting into, say, gain

18:28

site, for example.

18:29

So, there's absolutely the opportunity to, even before these direct integ

18:34

rations have been built,

18:36

to leverage those and push them into other platforms. I see, Emma, I see a

18:41

question from you,

18:42

what's the typical timeline for implementation for support logic, right?

18:49

And yet, so while it does depend somewhat on the complexity of your CRM

18:54

instance,

18:55

we can typically do it in 30 to 45 days, right? 30 days, assuming it's not

19:03

super complex.

19:04

And every CRM instance, every ticketing system is a little bit different, right

19:08

Every company I've worked for, like some of them, they have a legacy Salesforce

19:11

instance

19:12

that's more than 10 years old, that's massively complex. And some, it's a lot,

19:16

it's a bit newer, and they're hygiene's a bit better. But 30 to 45 days is what

19:24

I would typically

19:25

quote to be safe. Susan, good question. Thanks for the question about HubSpot.

19:33

Yeah, absolutely, right? So, our underlying architecture is Snowflake, is Snow

19:41

flake DP,

19:42

Snowflake is a Snowflake database, right? So, case data and case schemas, while

19:49

they're a little

19:49

different across distinct, across different ticketing systems, they're also

19:54

very, very similar. So,

19:57

as long as that data can be ingested via either ETL process or any connector

20:03

that you use,

20:04

which basically means any ticketing system, from any ticketing system,

20:08

we can get it into our Snowflake, and then we can map that data to surface

20:15

within our UI so that

20:16

it can be analyzed by our ML and our NLP to leverage some of these alerts in

20:22

this account

20:23

health. So, yes, absolutely. Our underlying architecture, it's CRM and ticket

20:30

ing system

20:31

agnostic. We can work with anything. Yeah, you're very welcome.

20:36

We covered so much already in some great questions. Let me show you another

20:43

thing, though.

20:44

Super valuable thing that I do use in support logic is I have a, basically,

20:51

I build a cohort of customers, right? So, this is our My Customers page. It

20:56

allows me to see a quick

20:58

snapshot of all of my accounts and their current support experience, allows me

21:02

to see the account

21:05

health, attention and sentiment scores, how many cases do they have open, like

21:09

how many escalations

21:10

do they have open, just gives me a quick glance at what's going on, right? What

21:15

's the distribution

21:16

of some of these, right? What's the age of them? This is super valuable just

21:20

for me to quickly check,

21:22

and I might do this. I don't do this every day, right? But I'll absolutely

21:26

check within the platform

21:28

and look at this view for my accounts once a week, right? The rest of the time,

21:32

I'm often relying on

21:33

the alerts to tell me when I need to either pop into our platform and take a

21:37

look or just to keep

21:38

me apprised of what's going on. And that's happening throughout the week on a

21:43

daily basis, right?

21:44

But you can configure your alerts to be really targeted. Obviously, if you're a

21:47

digital CSM,

21:49

right, and you manage 100 accounts or 150 accounts, you don't want to be

21:53

getting alerts for all of

21:54

those on a daily basis, right? So, you can configure those alerts for very

21:59

specific conditions. Maybe

22:01

you have some, maybe you have digital CS accounts that are coming up for

22:05

renewal, or that you're,

22:06

you know, working opportunities on or anything of that nature. The alerts allow

22:10

you to be very

22:10

focused and just be alerted in situations that matter.

22:17

Awesome. Any other questions?

22:23

Yeah, if we don't have any more questions, Max, thank you for this amazing tour

22:28

. It looks like

22:29

Steven's got a question on managing you. Yes, yes, yes, Steven, our product is

22:35

built from,

22:35

so the support product, the support logic product was built for managers. So,

22:42

this is

22:42

absolutely intended. These are intended as manager views, but what you can do

22:47

is if you're looking

22:48

to see a much, in a sense, a higher level view, right, I'm looking at

22:52

individual accounts here,

22:54

but because I create cohorts, right, if I'm, you know, if I'm a director of CS

22:59

or a VP of CS and I

23:01

want to see more of a roll-up, I can absolutely do that, right? You see, I

23:06

expanded this, for example,

23:08

but I could roll it up in this way and just have, okay, here's my, here's my AP

23:14

AC, here's my APAC

23:15

customers. Here's my, here's my Amia customers, here's my US customers, right?

23:21

Or here's my

23:21

customers that are using this particular product, right? You can build these

23:25

cohorts in any,

23:26

in any capacity that you like. And then you can also, obviously, we have

23:30

extensive analytics within,

23:32

within our product. Very much geared towards geared more towards support, but

23:38

also allowing you

23:39

to say, look, for example, right? Maybe I have, maybe across the different

23:47

products that my company

23:49

offers. There's been a spike, for example, in number of cases associated with

23:55

that. And like,

23:57

it could happen if my, if we had like a release or something, and I can search

24:01

for those types of

24:02

insights within, within support logic, but also be alerted, right? I didn't

24:08

talk about keyword

24:09

alerts, but you can absolutely create keyword alerts based on, say, your

24:13

competitor names,

24:14

right? So if your cohort of accounts, or, or any of your customers, your top

24:19

revenue customers,

24:20

or ever talking about a competitor in support, you can absolutely be alert in

24:24

that scenario as well.

24:25

Max, can you show Steven real quick that, that agent view at the top? He's in

24:30

customer support,

24:31

actually. And I think that view is powerful. Yeah. This one. Yeah. Yeah. Yeah.

24:36

So Steven, this is the,

24:37

this is the my agents view. And what I do with this is I can create a full

24:44

entry for my team,

24:45

but also for individual agents. And you'll see it's a heat map. And it's

24:50

basically looking at my

24:51

cases grouped by status, but also grouped by age, also grouped by an inactivity

24:58

. So I can know of

24:59

my team members, I can know which cases have actually been inactive, meaning no

25:04

inbound or

25:04

outbound communication for the, you know, for, for in this scenario, greater

25:08

than seven days.

25:10

And you see these are also color coded by the sentiment. So the ones that are

25:14

red, not only

25:15

of these cases been inactive, but the customer is not happy. And stale cases

25:20

are the bane of any

25:21

support teams existence, right? They clog up your backlog, make it, make it

25:25

seem like your,

25:26

your agents and engineers are like, have more, have less bandwidth than they

25:32

actually do.

25:33

So there's a ton of great insights that you can derive from that to use in a

25:38

tactical way as well.

X

This is a test comment.

X

This is a longer test comment to see how this looks if the person decides to ramble a bit. So they're rambling and rambling and then they even lorem ipsum.


author