Or they have the technology to automatically prioritise support requests based on the content of the ticket and customer value (eg whether you're already a customer or not). A support request from a non-paying customer asking a technical rather than a sales question might not be something they consider important, especially if their metrics should that those users rarely convert.
This was my first thought. There's no way this guy has a meaningful relationship/account with all of these companies. It's good business to prioritize where you spend your resources.
Without a doubt, and that's fine by me. Pay more, get better service. I don't pay much more than $100 for any one web service. Much less for most. And I know that some are one-person shows. I see a mixed bag. Some are not only very responsive, but also very personable and friendly! Others, not so much. Unfortunately, some one-person services are happy to bill themselves as such (indie!), but don't offer that personal touch in terms of support.
As one of the better performing companies in this study, I can attest that preserving the quality of our support as customer count grows is really hard.
But response time is just one, quite imperfect, metric. I wouldn't draw conclusions on that. We don't prioritize support based on customer $ value, and in my experience, customer support from our vendors never seems to be correlated to how much we spend!
As Nassim N. Taleb said, you wouldn't attempt to walk across a river that's on average only four feet deep.
If you look at the raw data (kindly provided at the bottom of the post), there are tremendous variations for individual companies between two email attempts.
For example:
Formdesk: 987, 6
Get Response: 4, 1445
Appointy: 12, 1209
Two attempts just aren't enough to reach conclusions.
But why do you post inconclusive data and draw potentially misleading conclusions, and then hedge with "I thought it was interesting and worth sharing"?
The argument of the GP is that, people are liable to draw the wrong conclusions. So perhaps the fact that the data are "interesting" is actually motivation not to post it.
It's categorically silly to withhold inconclusive data just because you don't have the time or ability to make final conclusions with it. This is a major issue with academic research today.
This isn't a question of transparency with regards to the data, but publicly drawing conclusions from statistically insignificant data.
Given the major issue of reproducibility in academic research, I'm not sure "withholding inconclusive data" is a major issue with academic research today.
While I find it interesting, I would be more interested in seeing the times of these companies when you are a paying customer. I expect faster responses, but I would not get surprised if some took ages..
Also the range of times the person requested support at, while limited, is still a broad timespan. For small or even some medium sized companies there could be huge variability in their response time.
I'm quite surprised how quickly people responded, I thought it would be longer.
In my business I commit to one working day, although issues don't usually wait that long I don't know how I could shorten it without having someone dedicated to support.
"I only emailed companies on a Monday or Tuesday morning during normal working hours (I don’t think it’s fair to expect customer support at midnight)."
It's not 100% clear whether the author means his local business hours or the company's.
IMO, it gives you an insight into the support culture of the company. If they treat support as "close as many tickets as fast as possible", then I'm gonna have a bad time.
If they treat support as "Make sure the issue is solved, and the customer only has to email once or twice", then I'm interested in doing business with you.
I've worked with two startups now, both with low support volumes, which grew many times upon acquisition. In both cases, these acquisitions brought upon an influx of new customers. It's simply bad business to scale your support staff based on MRR in the long term.
When the tipping point between growth and headcount arrives, a dev is usually assigned to figure out how to slow the incoming requests. There are many ways to do this, and some of the best companies have figured it out and mastered it very well at scale (see Google, Facebook, etc).
Some ideas:
1. Offer self-help/self-resolution documentation or tools (Gmail's self help is quite good!).
2. Categorize support requests, capture low hanging fruit, deploy fixes (sign up deficiencies, UI bugs, etc).
3. Set up a status page to proactively notify customers for outages/maintenances.
4. Increase customer communication when large functional or UI changes disrupt users.
5. Automation/AI/Bots (Not easy, and hardly ever done right)
6. Offer chat to increase interactions served by a single CS agent.
An interesting method, that is often heavily debated... Only offer support to paying customers. I think businesses can only get away with this, if they offer plenty of self-help documentation/tools, and your core product is relatively rock solid. Operating on a freemium model always brings about interesting challenges.
We implemented a 'smart' contact form. I basically was a decision tree based on our FAQ (updating our FAQ also updated our contact form). People had to choose the subject and some follow up questions wit radiobuttons. At the end (just 2 or 3 clicks) we showed them the answer from the FAQ. It that didn't answer their question they could fill in their question.
5^3 (5 choices and max 3 clicks) makes 125 possibilities and will cover most questions (if not all).
It made the support questions drop with 90% with the same customer satisfaction if not more.
TL;DR: People don't read FAQs, make them pick the correct subject for contacting you while showing answers to common questions. It will decrease support questions with 90% (n=1 though).
Link would be handy! :) Promise I won't accidentally send any messages.
The few times I've seen this done were pitiful, amazon's is easily the worst out of the bunch — but, I'd love to see a good example as it's something I'm tempted to integrate.
I think talking of average when the sample size is 2 is misleading. I find such a test (and possible conclusions) interesting, just wished it was carried out in a more rigorous way...
They should all be using http://www.leadactivate.com/ - it takes an inbound email and then places an inbound call to a phone number or multiple phone numbers and then uses text to speech to notify the agent about the call.
Typically used for sales, but it reduces response times down to about 20 seconds for inbound emails and leads.
For any given SaaS company, the fastest response is probably not the optimal result. At least 3 other variables are equally important:
1. Quality. How thorough is the response? As the blog post says, "Plus email reply time says nothing about the quality of support."
2. Cost. Are customers willing to pay more for speed and/or quality (either in dollars or because the company spends less elsewhere)? Up to what point?
3. Point of diminishing returns. Assuming a customer is willing to pay more, what's the real impact of a longer delay or lower quality? How {fast,good} do they actually benefit from?
Faster is only better when the test is only measuring speed. If you're reading this post and thinking "Man, our team should respond faster!" -- maybe they should, but maybe those other factors mean they shouldn't.
To give three examples I've seen in support teams:
- Company A has a big team of inexperienced staff sending fast responses. They succeed on speed (and would show up at the top of this list), but probably fail on quality. As a user, this is form letter hell.
- Company B has a big team of very skilled staff sending fast responses, or has developers or whole engineering team task-switch when a question comes in. They succeed on speed and quality but, depending on the product, probably fail on cost. May also have lower quality, since customer support/service is a skill that not everyone does well. As a user, this feels great. Sometimes the person who coded the feature I'm asking about is replying to me. OTOH, if I don't actually need knowledge that's unique to them, their time may be more valuable working on that feature :-)
- Company C succeeds on all of these, with a right-sized group providing high-quality answers quickly (let's say 2 hours, not instantly). Alas, customers aren't willing to pay for that. Their customers perceive diminishing returns from answers in less than 24 hours, and/or the price premium they want to pay for a faster (or maybe more thorough, or more clearly written..) answer is very low.
Sounds obvious, but based on how often companies make one of these mistakes, it's not. When one aspect is relatively easy to measure and others are hard or impossible, they tend to lose.
30 comments
[ 0.22 ms ] story [ 99.5 ms ] threadIf you signed up for a $5,000 / month plan with all of them, then emailed support, you'd have a different experience.
But response time is just one, quite imperfect, metric. I wouldn't draw conclusions on that. We don't prioritize support based on customer $ value, and in my experience, customer support from our vendors never seems to be correlated to how much we spend!
If you look at the raw data (kindly provided at the bottom of the post), there are tremendous variations for individual companies between two email attempts.
For example:
Formdesk: 987, 6
Get Response: 4, 1445
Appointy: 12, 1209
Two attempts just aren't enough to reach conclusions.
There's definitely a limit to what kind of conclusions you can draw from my data. Mostly I just thought it was interesting and worth sharing.
But why do you post inconclusive data and draw potentially misleading conclusions, and then hedge with "I thought it was interesting and worth sharing"?
The argument of the GP is that, people are liable to draw the wrong conclusions. So perhaps the fact that the data are "interesting" is actually motivation not to post it.
Transparency > statistical purity
Given the major issue of reproducibility in academic research, I'm not sure "withholding inconclusive data" is a major issue with academic research today.
In my business I commit to one working day, although issues don't usually wait that long I don't know how I could shorten it without having someone dedicated to support.
I wouldn't be surprised if overseas companies for example took longer to respond just because you reached them out of office times.
It's not 100% clear whether the author means his local business hours or the company's.
I'd really prefer to see the questions and depth of reply. Time to respond can be easily lower bounded by giving a shitty answer fast.
In my experience, customers are much happier with a response that takes longer, if they: 1) know it will come and 2) know it will be thorough.
IMO, it gives you an insight into the support culture of the company. If they treat support as "close as many tickets as fast as possible", then I'm gonna have a bad time.
If they treat support as "Make sure the issue is solved, and the customer only has to email once or twice", then I'm interested in doing business with you.
That is, if you know you won't be able to reply fast (weekend, late night, company party, etc...), let them know so you can set expectations.
When the tipping point between growth and headcount arrives, a dev is usually assigned to figure out how to slow the incoming requests. There are many ways to do this, and some of the best companies have figured it out and mastered it very well at scale (see Google, Facebook, etc).
Some ideas:
1. Offer self-help/self-resolution documentation or tools (Gmail's self help is quite good!).
2. Categorize support requests, capture low hanging fruit, deploy fixes (sign up deficiencies, UI bugs, etc).
3. Set up a status page to proactively notify customers for outages/maintenances.
4. Increase customer communication when large functional or UI changes disrupt users.
5. Automation/AI/Bots (Not easy, and hardly ever done right)
6. Offer chat to increase interactions served by a single CS agent.
An interesting method, that is often heavily debated... Only offer support to paying customers. I think businesses can only get away with this, if they offer plenty of self-help documentation/tools, and your core product is relatively rock solid. Operating on a freemium model always brings about interesting challenges.
Edit: Formatting.
5^3 (5 choices and max 3 clicks) makes 125 possibilities and will cover most questions (if not all).
It made the support questions drop with 90% with the same customer satisfaction if not more.
TL;DR: People don't read FAQs, make them pick the correct subject for contacting you while showing answers to common questions. It will decrease support questions with 90% (n=1 though).
The few times I've seen this done were pitiful, amazon's is easily the worst out of the bunch — but, I'd love to see a good example as it's something I'm tempted to integrate.
Typically used for sales, but it reduces response times down to about 20 seconds for inbound emails and leads.
1. Quality. How thorough is the response? As the blog post says, "Plus email reply time says nothing about the quality of support."
2. Cost. Are customers willing to pay more for speed and/or quality (either in dollars or because the company spends less elsewhere)? Up to what point?
Those 2 are the obvious trades (https://en.wikipedia.org/wiki/Project_management_triangle#.2...). There's at least 1 more:
3. Point of diminishing returns. Assuming a customer is willing to pay more, what's the real impact of a longer delay or lower quality? How {fast,good} do they actually benefit from?
Faster is only better when the test is only measuring speed. If you're reading this post and thinking "Man, our team should respond faster!" -- maybe they should, but maybe those other factors mean they shouldn't.
To give three examples I've seen in support teams:
- Company A has a big team of inexperienced staff sending fast responses. They succeed on speed (and would show up at the top of this list), but probably fail on quality. As a user, this is form letter hell.
- Company B has a big team of very skilled staff sending fast responses, or has developers or whole engineering team task-switch when a question comes in. They succeed on speed and quality but, depending on the product, probably fail on cost. May also have lower quality, since customer support/service is a skill that not everyone does well. As a user, this feels great. Sometimes the person who coded the feature I'm asking about is replying to me. OTOH, if I don't actually need knowledge that's unique to them, their time may be more valuable working on that feature :-)
- Company C succeeds on all of these, with a right-sized group providing high-quality answers quickly (let's say 2 hours, not instantly). Alas, customers aren't willing to pay for that. Their customers perceive diminishing returns from answers in less than 24 hours, and/or the price premium they want to pay for a faster (or maybe more thorough, or more clearly written..) answer is very low.
Sounds obvious, but based on how often companies make one of these mistakes, it's not. When one aspect is relatively easy to measure and others are hard or impossible, they tend to lose.