Customer service at startups

?

Episode 19: 25/09/2021

?

The owner of the reasonably expensive fish restaurant is standing at our table with the air of somebody who has to deliver some seriously bad news.

?

The menu had me a bit worried even before we sat down. It wasn’t the fish aspect. The prices were generally beyond the means of a self-employed tech founder. Particularly one who has been working on an un-launched, un-monetisable startup, for more than five years.

?

But we have not ended up at this restaurant out of choice. We are in the seaside town of Saint-Florent, Corsica, in late September. This is right at the end of the tourist season, and most of the other restaurants in the small harbour have already closed for winter. The evening is warm and there are plenty of tourists around, so the fact that the local businesses have already called time on the year’s activities is puzzling. Presumably their season is dictated by the school holidays, although I always do wonder if they wouldn’t just make more money by staying open later in the year and catering to the non-family market. Who knows, I’ve never dared to visit Corsica in November. Maybe the locals just don’t eat out.

Those other tourists are all making the same discovery as us, namely that they have only two options for dinner tonight: the one reasonably expensive fish restaurant, or the several mega expensive fish restaurants, which have also somehow managed to stay open. You can see the parents taking deep breaths as they shepherd their children inside for a night that will live long in the history of their bank accounts.

?

The proprietor of the reasonably expensive fish restaurant gravely informs us that many of the things we just ordered are not available any more. Tonight is the last night he will open until summer 2022, and a lot of ingredients are already out of stock.

?

We find this rather amusing. The original waiter went through the theatrics of taking down a long order for our party, making numerous mistakes and repeating everything several times, only for the owner to emerge ten minutes later and explain none of it was available anyway. Our smiles do not deter him. Eventually, his grim speech reaches its climax, where he lists which dishes are still feasible. It is all fantastic news for me, because one of those dishes is bread, which I can definitely still afford.

?

After he leaves our table, there follows a light-hearted conversation about his attitude. He was almost definitely playing it for laughs. At this point, readers familiar with this blog will be expecting some tissue thin link between the fish restaurant vignette and the saga of founding the El Toco search engine. Here it comes.

?

While munching on my fresh crusty bread, I spend the rest of the meal thinking about customer service.

?

Customer service

?

A few episodes ago, we discussed El Toco’s massive database of links. It was the stress tests of this database which, earlier in 2021, went so badly that they junked the computer we did them on, ruined our grad’s holiday, and were closely followed by their resignation.

?

The grad’s notice period was spent attempting to run the stress tests on the cloud. This, too, did not go well, culminating in an eventful last day, where the database server and all its backups got accidentally deleted.

?

To be fair, the fault for the backup deletion fiasco lies with Amazon. Their cloud servers are administered via a website. The website does not, anywhere, make it clear that if you delete a database server, the backups go with it. Amazon’s customer service confirmed this is "expected functionality". Er, what? If you accidentally delete a database, the backups get deleted too! Who thought that would be a good system!?

Fortunately, all the drama occurred in a test environment so nothing important was lost. But a certain sense of calm descended for all parties when that day drew to a close.

?

As of summer 2021, I have finished developing our new website back end so, while the designers works on the front end, I have picked up the thread of the database stress tests once more. The aim of these tests is to check how many links we can store in the database, before it runs out of space. I’m continuing the tests on the cloud, which is putting me in almost daily contact with Amazon customer service.

?

Good customer service makes the world a better place. Think about a time you received really good customer service. Mine is the homewares chain Lakeland.

?

I gave my girlfriend an electric blanket which stopped working after about eighteen months. That’s annoying, I thought, it’s just outside the statutory guarantee period for electrical products. Not with Lakeland. Their products are all guaranteed for two years as a matter of company policy. In one phone call, we arranged a replacement, which was actually an upgrade because the original electric blanket was discontinued, and it was delivered for free a week later.

?

?

The next time somebody tells you that there is no such thing as a free lunch, tell them about good customer service. Yes, we made the same observation a few entries ago, about speed at supermarket checkouts. Good customer service is another of those win-win scenarios. The customer gets whatever they wanted, and the provider is rewarded by repeat business.

?

The concept is really quite elementary, which makes it remarkable how often in life we come across companies who are yet to grasp it.

?

How to lose friends and alienate people

Everybody has war stories of bad customer service, but the ones from big business are more interesting to talk about. This is because, in a professional environment with large sums of money changing hands, you might expect bad customer service not to be a thing that happens.

It is very much still a thing that happens. From my limited experience in the world of big business, the best example is financial data providers.

?

Financial data providers deliver their data through some piece of software. They love to pretend that this software is the product, whereas in reality it is an obstruction, forever getting in the way of whatever you want to do. The software often has a moody black user interface, to make you feel like you're working in an episode of a crime scene investigation series. You install it on your computer and all that data, which is the actual product, flows in. Bingo, your newfound insights into the markets make you a millionaire.

That summary, while totally accurate, contains so many logical holes you couldn’t even use it to strain pasta. The big pasta, with meat inside. We shall ignore all those holes today, in favour of talking about customer service.

?

The software that delivers the financial data is extremely fiddly, to the point of being basically impossible to figure out on your own. It is also proprietary, so nobody talks about it online. As a new grad, I learned to my horror during week one that you can’t just look up how to use financial software on YouTube or Stack Overflow.

?

Users of financial data software have absolutely no choice but to rely on the provider’s after-sales support. Can you see where we’re headed with this yet?

Yup, you guessed it, the after-sales support sucks. Reps swarm your office, with the remarkable ability to show up just when you’re deeply concentrating on something or on the point of going for lunch. The reps are well-versed in the software, but this is irrelevant. They are merely to provide face time and are too busy visiting clients to help with anything beyond basic queries. If you have a question, you need to go to the Support Desk.

?

The Support Desk is not so-named because it provides support. It is called the Support Desk because you need support after you’ve finished dealing with it. Every single wretched time you have to contact the support desk, the story plays out as follows.

?

You email the support desk to ask your question. The support ticket is assigned to a new hire, straight out of university, who doesn’t know how to answer it. They are merely a filter, to probe topics like “is the customer’s computer switched on”, "is the customer a customer", and so on. Meanwhile assessing how much the customer really cares about the issue, in case it can just be left unsolved.

?

After an interval of between hours and days, they reply. That delay is a key part of the filtering process. People at both ends of the priority spectrum will have fallen off the radar by that point, having either got bored of waiting or been fired in the interrim.

The first reply from the support desk will always be incorrect, and only tangentially relevant to the thing which you asked. As soon as it arrives, you don't need to read it. You must immediately bounce back with a rephrased version of your original query.

?

Thus begins a game of ping pong. The support agent sends replies which don’t solve the issue. Each time, you respond by rephrasing your original question. This table tennis continues until you get bored of the topic or get seriously shirty. A response that always got me from nought to get seriously shirty in two seconds was whenever a support agent patronisingly told me that our in-house developers should answer my query, when I was one of those in-house developers.

?

When the customer gets seriously shirty, they have passed the initial filter. Their support ticket then goes the second stage.

The second stage is a lottery draw, containing you and all the other unhappy shoppers whose support tickets who have reached this point. Once in a while, the handful of people at the provider’s end with the knowhow to answer these questions dip into this lottery, pick a ticket, and the lucky winner has their problem solved.

You will have typically been emailing for a fortnight or two by the time the experts are put on the case. Generally, one single email from these people is all that it takes to solve your issue.

Generally. Occasionally, though, you will have discovered a genuine bug or problem with the software. In which case this, too, will degenerate into a second round of ping pong, where you email back and forth with the domain expert, repeatedly explaining that Feature X is doing the wrong thing, while they try to pacify you without actually solving it.

The second round of ping pong ends either when the bug is admitted, or the domain expert stops replying to emails. Notice that, if you get to this point, the bug not actually going to be fixed. You're not important enough. Numerous customers need to reach this stage, and still care about Feature X working correctly, before it makes it onto the development roadmap. For the following year.

?

?

The six years I worked as a quantitative analyst provided a steady stream of low-grade comedy for my colleagues, as I pursued support tickets to extents where they would have either found a workaround or just abandoned the matter entirely. They particularly enjoyed eavesdropping on support desk phone calls, which normally went like:

“So I have to click on this button here do I? The one that’s greyed out, because it’s disabled? Yes, I’ve tried restarting the software. Yes, I’ve restarted my computer. Yes, we've reinstalled the software. No, the button is still greyed out.”

?

Given this trauma, I was prepared for the worst when, five years later, I started regular interactions with Amazon customer support. Amazon's cloud is significantly more complicated than a Bloomberg terminal, and Amazon servers bill you by the hour, even if they aren't working properly.

Fortunately, they put the financial data companies to shame.

?

?

?

Amazon customer service

Amazon Web Service customer support is how you would expect customer service to work, if you were an alien arriving on Earth with no experience of how customer service normally happens.

?

You have a problem with an Amazon service. You contact Amazon. Amazon solves the problem.

?

We have talked numerous times about the astronomical value of time for El Toco. While we are pre-revenue, bankruptcy is always a shadow on the horizon. The things we use need to work out of the box and, if they go wrong, somebody needs to fix them with a minimum of effort from our side. Amazon’s good customer service is a real boon for us, because whoever set it up understood all this.

The support agents are almost never in our timezone. It doesn't matter. We often ask stupid questions. They don't care. We are also very thorough, so our questions are quite exhaustive. They always get answered.

Tiny as El Toco is, we have occasionally spotted genuine bugs, and Amazon have, in every case, managed to find workarounds, so that the bugs didn't turn into showstoppers.

?

The whole experience is so refreshing it makes you realise just quite how bad the customer service is in other places. I wasn’t involved in the commercial side at Barings, but rumour had it that some of our financial data bills ran to the hundreds of thousands of pounds annually. While we are not launched, El Toco’s annual Amazon spend is about ￡1.5k. What on earth are the financial data companies doing, that their customer service is so bad, while we pay Amazon ￡29 a month and they can solve all our queries?

This is not to say Amazon is perfect. Timekeeping is still an issue. The instant something breaks on our cloud, irrespective of what it is, I know that another few weeks have been added to our launch date. This is the time it takes on average for support tickets to be solved.

Also, at times, we've had to be quite firm with Amazon. The rough and tumble normally happens when an Amazon service is interacting with a third party product, like Microsoft Windows.

Suddenly, hands are thrown in the air and we're back in computer-says-no-land, as they invariably lay the blame on the other company. The trouble is, as a micro startup, we aren't big enough to merit customer support from the other products we use. Microsoft aint gonna help, and open source stuff like Linux doesn't even have customer support. If something breaks, Amazon have basically got to solve it, because nobody else is going to.

As a result, we've got good at wording support tickets in a way that emphasises the fault is with Amazon. Then, the customer support agent doesn't try to wriggle out of answering the question. I'm far too inexperienced to give actual advice about using the cloud, but in terms of managing support tickets, this has been a skill worth mastering.

?

Wrapping up for launch

Even with Amazon's best efforts, the stress tests of El Toco's links database have still gobbled up two months of my undivided attention. But, at least, this has been due to the difficulty of the task.

?

The conclusion, after some painstaking experiments, is that our links database can hold 100 million addresses before it grinds to a halt.

This is obviously a tiny fraction of the internet. But it is enough for our budget. Beyond it, we will need a plan to scale things up. We will cross that bridge when we come to it.

??

We’ve spent the better part of 2021 doing the advertising section of the website. Now we are back where I thought we were at the end of 2020: on the cusp of launch. The advertising section all works, the links database looks healthy, there's not much left on the todo list.

?

There's just one thing left, actually. It is to finish polishing the visuals for our website.

The website company is working on it at the moment. We’ve had some problems with this contractor in the past. They are working under a much more stringent contract this time, ensuring less scope for slippage. As the graphic below suggests, this won't help.