Peter Schlegel, CEO, Admazely
Presentation at SlideShare: 100% monthly growth: How we solved the most important problems.
January 29, 2015
So actually, the story that I’ve been telling over the past shy of two years is how we went bankrupt – not that that’s much of an accomplishment. But apparently I’ve become a very popular guide to talk about fucking shit up, so I’m pretty excited to talk about some of the things that we’ve did right while we were fucking other stuff up. I’m very happy to be here – just wanted to say that.
Briefly about Amazedly, what we did is that we’re a retargeting company. Retargeting works like a consumer visits a webshop and leaves the webshop, and then you can retarget using ads and make the consumer come back and buy the stuff that people are still interested in. Like ad stalking, basically – we’ve all seen it.
What we did was that we’ve found a pretty clever way of doing that where someone had a webshop and could use our technology to crawl and scrape the entire site for products, then we applied some very clever machine learning and then we would give them a UI with all the products and some unique banners, which was, at that point in time, was a service that no one else was really offering. Briefly, that’s what we did. We sold that stuff to webshops, which is far more interesting actually for this case rather than what we did specifically.
This is a timeline with our active customer base. We did a lot of R&D here; we didn’t have any, and then we sort of gotten the first few customers, and we raised some money around – I think around here – a seed round of about half a million. Then we had to get some more money because we would run out in a year, and we were very confident that we could do that because it had been very easy to raise the first round of the money.
It turned out to be not quite as easy to raise this second amount of money. We were fundraising here when we had some very little and then we were told in February, before we sort of had February figures, “Sorry guys, we’re not giving you any money,” and then we did this – which was a bit of a shame. Then we went bankrupt up there. That was a bit disappointing, I have to say.
But nevertheless, my first lesson here is not for growth hackers but to founders, so you might want to gain traction before you run to raise money, which is obviously easier said than done, but nevertheless.
What I want to talk about actually is more like the things that we did to gain this momentum in these last three months, rather than the miserable story of how we went bankrupt and I got all depressed and stuff.
This is our sales funnel, and we have some missing conversion rates. We’d just talk about from the beginning out here. Ideally, we wanted recurring revenue – that’s pretty much what you want as an enterprise startup, especially one that is a Software as a Service.
The way that we did this was basically we would call people up, we’d talk to them then we would give them a demo, and they would agree to the pilot or trial project, and then they would become happy and become recurring revenue. And we had some conversion rates around that.
The investor that we had had some pretty fresh memory from TrustPilot, which had the same basic model, so we had some pretty good benchmarks, and we were very fortunate to talk to other people who have been doing this telesales model. Understanding how a performing competitor, a very successful company, was pretty helpful in trying to sort of find out whether we need to target these different things.
I’ll come back to some of these, but we’d just start by saying that these are the percentages, but the main thing that the people told us was that the volume of calls made per sales rep was extremely low, extremely inefficient, so we ended up figuring out what we really needed to optimize. For me, at least, that was the big growth hack that we did, so I’ll save that for last and leave you all in excitement for now.
But – let’s go back one. The growth hacker lesson here is to try to quantify this from day one, very early on, even if you have very limited data volume. Try to look at data, try to be very data-driven or insights-driven in all of this. Measure from day one, even when you’re only doing 10 calls/day, still figure out how successful are you. Otherwise, you won’t know if you’re improving, you won’t know if you have a problem, so try and do that.
A lot of the people that we talked to, like TrustPilot for instance, are saying, “Yeah, so we have a salesforce, then we have these very cool integrations with other technologies that will then measure automatically how many calls a sales rep does per hour, in a day, and spit out a lot of chats.” Then I looked at the license fee for salesforce and we kind of realized that was probably not something that we wanted to do.
What we were doing at the time was that we bought some prepaid Skype accounts and then we just call from that. We started using the free stuff, which would fall out all the time, which again, didn’t really do much for our conversion rate.
So we bought some paid ones at 100, 200 kroner at a time, and we were calling –. To quantify this, we would extract the Skype logs for each of the sales rep, manually them. And then we’d put it in Excel and we would manipulate and say, we may have a call, but if it’s less than 90 seconds, then it’s not really a conversation. It’s probably just calling the reception, or the phone was busy, or we couldn’t reach the right person.
At first, we just sort of had a conversion rate, call to demo, that turned out to be a little bit meaningless. We had the sales reps, that being me included, register manually in this ZoHo CRM, which was at $25/month, so that was pretty manageable.
And then we would consolidate all these things, and we can measure of course how many customers we had by the amount of invoices that we, which we use the e-conomic, which I’m sure you all know.
So seriously, quantify and get the different data sources together very early on so you have an idea what’s going on.
Now, a few actual hacks.
The first thing that we didn’t really do was solve this whole pile to move the current revenue, because we didn’t deliver enough value, which was, in many cases, related to the fact that we had no idea what we were doing with the actual retargeting. We just saw some companies who were catering to very big companies be very successful, and we figured that it can’t be very hard to do that. The trick is the stuff that makes it work with the machine running for the small customers, the whole.
My initial assumption was, “Surely, this is as easy as scratching my own ass” to actually make this work, and then I went “Fuck, that’s not so much.” So I had to go into this system that we’re using that we’re using to actually traffic the ads, very similar to Facebook advertising interface and so forth. And then also I’m spending a lot of time in this interface looking at reports, and it’s very efficient because then I have 10 or 20 campaigns running, and if I segment them, it explodes, so it was not very scalable.
We tried to do some nice, automated data extraction to Excel and we did shitloads of pivot tables on a lot of different parameters, and we learned a lot from that. And then at one point in time, we also said, “Actually we got a PhD machine learning and a theoretical physicist in our team, maybe they can help us with this whole data thing,” which obviously, they could, but by the time we got to this realization, we were very close to bankruptcy, so we really harvested the full benefits of this and they were busy doing other stuff.
The next thing, like the Demo to Pilot – this is where we were performing extremely well. The conversion rate of everyone we actually got to talk to and show the product, then six and ten says, “That sounds awesome; I’d like to pay you some money to try that out” were very happy with that. Compared to the early days of TrustPilot, we were really even kicking their ass on this. Obviously not on every other single parameter in the business, but for this one.
But the product didn’t quite work; it sort of didn’t really do what we wanted it to do yet, so we had to do a product hack for this. One of our developers, he saw that we got two customers that were using Shopify as their webshop platform, and he noticed that actually the site map was constructed in the exact, same way. He looked into it and found out that there were some similarities because it’s like a standard product, and that meant that it was done in the exact same way.
All of the stuff that made the product not really every website in the world didn’t really apply to every Shopify webshop in the world, so every Shopify webshop in the world, we get to hack, and the product would work immediately with all those.
So we went to BuiltWith – an awesome website – builtwith.com, and then we used that to find all the Shopify shops that we could come across. And then what we could do is that we could then repopulate our demos with the actual customers’ own shop and own products, and that meant that the whole demo to pilot conversion was hell of a lot better than even the very good 60%. So that was very uplifting.
That kind of got us started with a lot of other ideas as well. For instance, like the process that we had with Shopify shops was on every other aspect very similar to what we had with any old shop built on anything. But with Shopify, we’ve automated, so we started automating a lot of stuff and tried to be more market-centric with this, so didn’t need the whole handholding, but we never really got to the bottom of that.
But as Kaare was saying just now – I think that Matias probably also had one in his points with Facebook – is that try to find niche audiences that you can optimize for. Try to find these thin slices in the market that you have something special to offer, and then offer that to those. Because regardless, especially if you were there in the early days, it’s a really good platform for learning a lot of stuff and maybe also to harness a graph, which obviously makes it easier if you’re a VC-backed company and you need to raise money.
Yeah, I just said that.
And then the whole calls-conversations ratio was also – that was not really a hack; that was just trying to coach the people that were doing it, giving them some sales training and all of these different things, so no magic to that.
But the big thing was this is where we were really wasting time and money in this process, and we didn’t really have an idea solely because we look at the actual numbers, like how many calls per day does the sales rep perform, and then we were shitty at that – extremely shitty. We spent a lot of time talking about it, and this is sort of what our idiocy in this looked like.
We had some segments. We wanted to target small to medium-sized webshops, we wanted to talk to these sort of owner/manager or the marketing manager and wanted them in the UK. Process-wise, we would go out and buy a Dun & Bradstreet database – and you can’t buy webshops from Dun & Bradstreet database. You can buy companies with revenue between x and y, so we bought some of those, and then we sat down and then we had people find these companies online, visit their website, trying to figure out if they actual sold something online. In most cases, they didn’t, so we couldn’t really use that lead at all, and then we had just spent anywhere between 10 and 20 million, just figuring that out. Maybe not 20, but too much time at least to do on thousands and thousands of leads.
And then if they didn’t have a website, then we needed to find the telephone number, then we need to call the reception, ask for the name of the guy in-charge of this and then try to convince the receptionist to transfer us, which is a hell of a lot more difficult in the UK than in Scandinavia. in their structures. So this was an extremely inefficient process that we had going here.
And like the last time, we had one of these events – I think all of the speakers tried to come up with a definition of what is growth hacking, and. To me, growth hacking is something about combining computer skills and sales and marketing skills to apply non-linear problem solving to otherwise linear problems. So it’s not just about optimizing little things that are obvious; it’s about trying to do something that can actually make a difference that is an order of magnitude, and I think that’s a part we ended up actually achieving.
So the new process that we defined a segment, so we wanted to target fashion webshops, for instance. Then we would brainstorm within fashion what are the 250 search terms that people will use when finding a website within fashion, like women’s jeans, or men’s shirts – all these different types of things. Then we would use Google Keyword Planner and we would type all these in Google Keyword Planner and then that would become thousands of keyword searches that were related. It’s a very good way of sort of scaling our own ideas for finding potential leads in all this.
Then we wrote a robot that was a crawler – not this one, but I couldn’t really find one that we did. You all know what a crawler or a robot is, so I’m not going to spend time on that either.
So the bot would then type in search terms – all of these thousands and thousands of search terms, one at a time, into Google, and it would get results on thousands of Google searches, and then we would look at the Adwords. Who are advertising, searching for –? Who are willing to pay money for people looking at these types of products? Obviously, they had something to sell and they had a marketing budget already there. We are having far more qualified people that we could call, so this way, we would get tens of thousands of leads into our database.
Then the bot would click the site or click the ad, visit the site, would look at the source code, search for “basket” or “cart” – the things that indicate that these are webshops instead of just random sites that we couldn’t do anything for. And then we’d get a list of tens of thousands of pre-qualified leads in this sense.
Then the bot again would take all these things and then hammer it into Alexa, which is a traffic estimating tool for the different sites, and we kind of use this to also find our sweet spot. There are some sites where we’re too large and some are too small, but this way, we would get an Alexa rank for all of these different ones and pretty fast we knew how to disregard the ones that were too big or too small. This way, we would get a list of again, tens of thousands of very, very qualified leads.
Then we would give them to interns. I think this is one of the beautiful things about an Anglo-Saxon country is they have a very, very beautiful tradition of doing unpaid internships – got to love that in a startup. So we had a few interns, we’d give these to them, then they would call the reception for info. They didn’t want to try and get transferred to the right person; they would just say, “Hi, I am just updating my database. Just wanted to make sure that Jenny Smith is still doing online marketing in your company.” “No, no, we never had anyone called Jenny.” “Oh, so what should I put in my database then?” Then they would get the name of that person. And then they would hand over to sales, who would then give them an actual sales call.
This was a new process of doing these things, and I think it’s fair to say that we sort of hacked our way to this optimization. These were our results before interest, taxes, depreciations, amortizations and bankruptcy.
I mean, we did a lot of stuff wrong, as I said, but this actually worked out extremely well for us. I think the thinking that goes into this, I think, is really trying to find us – as I said again, if you’re a software company, you have programmers. Use those programmers to tackle sales and marketing problems. That’s really what defines growth hacking; otherwise, you’re just doing sales and marketing and applying marketing skills to your own job title, really.
Some lessons learned for us, in this case. Work very much with this framework: hypothesis, experiment, data, insight, outcome. Be very hypothesis-driven. We’re very lucky that we had the sales funnel and we had some input into where we were doing well, where we were doing shitty, so we could develop some quick, good hypothesis around that, which worked out really good for us.
There are reasons why they test on guinea pigs before they test on humans. Find guinea pigs. Find areas where you can test without really harming anything. Do cheap tests on places where you can afford to fail; don’t try to do a very big, expensive experiment because if that fails, then you’re probably pretty fucked anyway. So do these things small on guinea pigs before you roll your stuff out on humans.
Leverage others to hack your learning curve. Look for people with domain knowledge. I’ve said it quite a few times, like in our case, there were Danish startups who were very good at this enterprise telesales model, and we talked to a lot of them to get some input on how to do these things, and understanding how we should train them, what kind of skill set should we look for, what were the challenges that we’re going to encounter. Do that, talk to people – not just to hang out with other cool startup people, but to actually learn something from people, to know something about the problem that you’re attacking.
And apply dev resources to sales and marketing. There’s a reason why software is in the world – it’s because it allows us to do a shitload of things more efficiently than otherwise. Especially if you’re a software startup, by definition you have these resources. Channel some of that into your sales and marketing efforts and everything else is just a big, fat waste in your development product and no one’s going to buy it because you don’t market it well.
And then learn to love the problem-solving process, not the specific solutions. Don’t fall in love with wanting to solve a particular problem in one particular way. From our own experience, I said that we got really clever about this whole finding-a-special-solution for Shopify. The truth is that I think for six months, everyone we talked to would say, “Well can you do something for a specific niche, maybe a specific platform?” And we said, “No, no, no.” Once you do something that solves the entire generic, global problem at once, that’s what we’re driven by, which was totally idiotic. We should be driven by the problem-solving, not the specific solutions.
Don’t kill your talents. Don’t get married to specific ideas on how to solve a problem. Be very open and intuitive about the problem-solving itself.
Lastly, what I wanted to say and I thought he was going to be here today. I’d ask him to stand up and point at him. I’ve said a lot of “we” in this process, and when I’ve said “we,” nine times out of ten, I meant a guy called Daniel Tronier, who was doing all of this stuff. He’s an amazing guy. I’m not doing a startup right now, so I would recommend anyone looking for someone to help you with this growth hacking, to grab hold of him. He’s a really kickass guy.
If you have anything, feel free to get in touch. Twitter handle, email and phone number. I’m very happy to answer any and all questions that you have in this context. Thanks a lot for your time.