Ask HN: Do bootstrapped businesses lose much money on free trials? | Hacker News Reader

Ask HN: Do bootstrapped businesses lose much money on free trials?

44 pointsfreedrull30 comments

I assume these companies probably run on ec2: https://www.hostedgraphite.com/ https://circleci.com

And probably many more that I can't think of at the moment...

I noticed many companies that do something similar offer a free trial for their services. How much money do you think they lose out this way? I assume they probably run an ec2 instance per customer, at least for the hosted graphite.

The cheapest ec2 instance is around $15/mo. If the trial is 30 days I’d assume they’re likely losing a lot of $15's here and there.

Am I wrong?

View on HN

This is one of those "How long is a piece of string?" questions. It depends on each individual business' circumstances. The vast majority of SaaS businesses run in a multitenant fashion and don't have meaningful hosting-based COGS for a marginal trial. (I know those two examples you picked are not representative of the typical SaaS, but I'm going to take this in another direction anyhow.)

This might be news to much of the peanut gallery, so here we go: COGS (cost-of-goods-sold: servers, etc) is not nearly the dominant hard cost associated with free trials. It is, overwhelmingly, cost of customer acquisition -- i.e. the marketing spend required to attract one customer at the margin. B2B SaaS companies routinely pay hundreds of dollars to get one new account at the typical $29 to $249 monthly price points.

Why do we do this? Because of the SaaS LTV formula: monthly price / churn rate. If you have a mediocre-ish 5% churn, then a $30 a month account is worth $600. If you can buy them for $200 apiece, why not do so? ("Massive stress-inducing cashflow crunches!" Good answer, and that's the classic reason why SaaS companies raise money despite having near-zero COGS and products which can be bootstrapped to market.)

$200 per new account businesses might imply, in a card-not-required-upfront model, 2% conversion rates on $4 trials which have converted at 25% from $1 clicks. Or, in a card-required-upfront-model, 25% conversion rate from trial, $10+ clicks.

To elaborate on patio11's point: I actually do spin up an ec2 instance per customer, and even then it's not really a relevant cost compared to the cost of acquiring customers.

Why would I do such a wasteful thing? It's totally not worth the engineering effort right now to switch to a more efficient multitenant architecture. Even a single new customer is worth more to me than I'd save in a year.

Twenty-first century rule of thumb: If the single-tenant architecture is the thing that you build first, because it's less effort, you should probably never bother with multi-tenant, unless a truly compelling rationale appears. VPSes are cheap and they will keep getting cheaper.

Multi-tenant is such a pain. Your customers become coupled on so many axes: Security, performance, downtime, resource management, versions of software in production, retirement of obsolete hardware. Everything one customer does effects others; you can't touch one customer's setup without potentially affecting others; if you try and coordinate the customers via email or phone, deadlocks will happen and progress will grind to a halt.

I worked on a product that launched multi-tenant and swiftly pivoted to single-tenant; it saved our bacon. Before that the support team was working overtime issuing lame apologies: "Uh, sorry about your lost sales due to slow performance; your neighbor has no idea how SQL works."

Counter point to this was a recent client who is working to consolidate the seperate VPS' because the combined data of different clients is so much more valuable - insights About customer data even when gleaned from competitors is useful.

(Talking about terabytes of usage data in aggregate here btw). YMMV

Any rule of thumb as general as the one I proposed is bound to have exceptions. For one, when the tenants know each other and have formed a consortium to allow their data to be aggregated, it's not the kind of "multi-tenant" I had in mind. Now, instead of a bunch of independent renters, we have a condo association.

My first instinct in a situation like this would be to ask: "Why not build a separate data warehouse and have some worker processes copy the clients' usage data over there, in batches or in a stream, raw or pre-processed as you prefer? Then the data can be be indexed however you want and JOINed with wild abandon, without further perturbing the individual client systems and their existing, presumably working architecture." Isn't that the old-school IT way? Yeah, I know, terabytes, but a terabyte isn't as scary as it used to be. Of course, the devil is in the details and I'm sure there are reasons.

To be fair those were some of the to and fro arguments we had - mostly I think it came down to the pain of data transfer (by then it had moved three disks iirc). A Hadoop cluster was being built when I left. I guess I should catch up and see how it's going.

But yes - definitely not the typical SaaS use case

I seem to be struggling with the maths in this

  2 cust from 100 trials (each trial costing 4 USD) where the 100 trials came fron 400 clicks (PPc of 1 USD)

This implies we spent 400 + 400 USD to get two customers or a cost of acquisition of 400USD

The other card-required one seems to be

  25 customers from 100 trials (cost if trial still presumably 4usd) 100 trials come from 400(?) clicks at 10USD

So that's 4400 to get 25 customers or 176 USD per customer (much closer to 200

I am not nit-picking for the same of it - just trying to follow what seems to be something deeply ingrained in saas owners psyches.

Edit:

Just in case it's obvious I want to say it has been a long day and I have a cold :-)

Possible that I botched mental math at 3 AM, let me check if after morning coffee.

My own coffee is still a commute away but pbiggar reminded me of some similar comments from yourself

600 LTV - but don't spend more than 1/3 on acquisition (200)

Assuming 4% trial to cust conversion we need 25 trials to get one customer. At 4 USD per trial that's 100USD

Assuming 1% click to trial conversion we need 2500 clicks to get 25 trials. That's expensive. So we need much better click to trial conversions. Assuming 1USD per click we need a 25% click to trial conversion.

I think it feels more likely to swap those rates around (4% click to trial, 25% trial to customer) but whatever my intuition is still waaaay out.

But either way it surprises me how tight 200 USD customer acquisition is. It sounds pretty close to the steak dinner high touch sales you mention (I can think of some very nice places in London to lunch a client for 200 bucks) but in reality it's eaten away by conversion rates far far faster than my intuition allows.

Thank you - whatever the "real" figures are working this through helps align my intuition with Mother Nature and if you had not put the numbers up I would not have worked it through - and when my billion dollar SaaS launches I would not have had the right intuition - which is the whole point of the exercise !

I miss excel :-)

If you're getting single digit click-to-trial conversion rates in a low-touch model with no card required, you're doing things wrong. 25% is a benchmark and people who are really good with ad targeting and squeeze page design can get 40%+.

(This is a major reason why, if you actually have a high-touch model or your software isn't of the kind that someone clicking an ad can reasonably buy on the spot, you will attempt to sell them on "Give me your email and I'll give you something you want" rather than "Give me five minutes and I'll persuade you to try this software.")

I have a lot of data on landing page performance but, unfortunately, most of it is confidential. Bingo Card Creator is the easiest one for me to talk about, and that's almost locked at 25% squeeze page to trial signup for several years running.

Apologies for the delay -- been a long day.

I re-endorse this math, with the elaboration that $10+ clicks was me being handwavy about $12.50.

I was not challenging the experience behind the maths, just I don't quite follow the formula

I think that the above means 2 out of 100 trialists will become customers and 1 out of 4 people who click on a ad will become trialists. So with 200 clicks we will get 50 trials and then one payin customer

If the trial costs us 4 USD in hosting fees that means we pay 200 bucks (50x4) for the trials, and 200 bucks for each 1 dollar click through (if that's how click throughs work).

So that seems to be 400 bucks to acquire a customer. This could be changed in so many places (cost of hosting etc) that the actual figures aren't really important - just trying to follow the rough proportions

As you can tell my uncertainty on how click throughs work indicate I am trying to understand the landscape - not challenge your experiences.

I am however very surprised that click to "free trial" is such a big conversion (25%) but after that they can't be bothered to go further. My takeaway from this, completely at odds with what I would have guessed last week, is to always take a credit card at sign up. Anyone who can be bothered to open the wallet is probably really interested.

And if they don't open the wallet, grab that email :-)

Tl;dr I believe you. I just don't quite follow the details - and that's where the devil lies I am told.

We ran the numbers on this last month, for our SaaS startup (shameless plug at end of this post). Certainly Patrick is right that the marginal cost of an additional user is low. It's why the math works on a "free for life" entry level plan. But in our case, we offer SMS (including international) on paid plans, so a free trial has a real non-zero cost for us. And there's an additional risk attached to offering a free trial and ending up with a riptide of unprofitable customers.

Of course, not having a free trial could also have a non-zero cost. We made some assumptions and played with a model but in the end we followed our gut. We have that free for life plan, and an affordable, under $10/mo entry-level plan. We offer free trials to customers when it seems appropriate, and we'd certainly give one out if asked, but it's a manual process for us and we like that.

* Cronitor offers dead-simple cron job monitoring and time tracking. Check it out at https://cronitor.io

FWIW, offering free trials in my SaaS business was THE strategic decision that jump-started our growth. Before that, I would do a screen share (or worse, a guided walkthrough) on the phone with the potential customer and it just wasn't convincing enough, alone, to get them to sign.

Once I started offering a 12-month contract with the first month (or 2 or 3) free + cancellable, that's when I started getting paying customers. That period of usage allowed them to begin trusting us beyond a short pitch.

The important part is to get them into a contract for a full year with the free trial period tacked on the front. If you just do the free part with no contract, they probably won't be proactive enough to sign up, even if they liked it. What you want is for them to be lazy after the free trial period is done, not cancel, and become a regular paying customer.

And yes, this is very dependent on your costs. My business was like cable TV: each additional customer brought zero extra variable cost and 100% extra revenue. So the free trial strategy was very, very strong.

Would you track usage after first month and prompt or cancel those who clearly had abandoned and were not using it but still being charged?

I'm a founder of one of the examples you list: https://circleci.com

Most of the people on this thread have it right: acquiring customers costs money but you make it up in the long run. You should be doing your SaaS metrics and include the cost of trials in your CAC, which should of course be less than 1/3 of your LTV.

Because we're an infrastructure-based business, there is a bit of cost in giving people a trial, but it's not really that big a deal, and free trials are largely noise considering our overall run-rate (in January, we announced that this was over $1m ARR).

And of course, raising money makes this easier too: we've raised about $7.5m at this point (http://blog.circleci.com/we-raised-a-6m-series-a/), so even if our trial was expensive it wouldn't break the bank.

I don't have any knowledge of the inner-workings of either company you listed, but I can tell you from experience its highly unlikely they're running an EC2 instance per customer. That would be very expensive and very weird.

its highly unlikely they're running an EC2 instance per customer. That would be very expensive and very weird.

Not so weird: https://www.mongohq.com/pricing/dedicated

Their business revolves around supplying dedicated database servers...

It would be a very weird architecture for the average SaaS product.

The existence of a company that offers dedicated EC2 instances to their highest tier of customers (and no free trial on that tier FWIW) doesn't meaningfully challenge the GPs assertion.

Most bootstrapped SaaS I know (including mine) do not, indeed, run a dedicated EC2 instance per customer. Some I know do, but it is in the "dedicated / contact us for pricing" option, and the price of the plan makes the EC2 price a detail.

As you say, we don't (founder of https://circleci.com here). Our infrastructure runs on LXC, so we divide up big EC2 machines into about 10-12 containers, and of course it's multi-tenant so if you're not pushing code you're not costing us money.

The cheapest ec2 instance is around $15/mo

The on-demand t2.micro is $0.013 per hour, so in a 31 day month, you would spend $9.69. The t1.micro is available as spot instances, and currently the price is $0.0061 per hour (fluctuates with demand), so if you were able to get it at that price for a whole month, you could pay $4.53.

I doubt they are spinning up an instance for each trial, but let's say they are. If you get 100 free trials and it costs $500 to $1000, if you can get 10% to sign up for your cheapest plan ($19/mo) and the average customer stays with you for 3 years, you'd have $6840 of revenue after 3 years.

"Do bootstrapped businesses lose much money on free trials?"

No, they don't lose money. If a business is at a net loss offering free trials, then they'd stop offering them.

--- start example ---

Let's suppose your yearly revenue from an average customer is $96 ($8/mo). And suppose your free trial is 30-days ($8 value per trial). What conversion rate from trial-to-paying do you need to rationalize offering a free trial?

Well, 10 trials is a cost of $80, so we would need one customer (+$96) per 10 trials (-$80) to convert in order to make the trial offering a net-gain. If your conversion rate is less than 10% then it wouldn't make sense to offer free trials.

Now, this is all assuming that a customer wouldn't sign up straight-away if you didn't have a trial; which is not necessarily the case!

--- end example ---

The goal of a free trial offer is to convert the trial customer into a paying customer and the cost of non-converted trials is factored into the cost of customer acquisition.

Every service is going to have their own implementation - at OwnOcean we spin up a Digital Ocean instance for each customer so that each customer has their own space. So with https://www.OwnOcean.com, we don't have trials but we'll happily give you a refund. It's simpler than offering free-trials and functionally equivalent.

Bypassing the question about costs, even if they would be "likely losing a lot of $15's here and there" then such costs are often considered acceptable for customer acquisition.

For example, back when Paypal was getting started, it simply gifted $10 to every new user, and in B2B sales you can expect to spend thousands on every prospective customer, even if no sale happens in the end.

Looks like hostedgraphite is on Hetzner, while circleci is ec2:

    [localhost Downloads]$ whois `dig +short hostedgraphite.com | head -1` | grep -i netname

    netname:        HETZNER-RZ13

    [localhost Downloads]$ whois `dig +short circleci.com | head -1` | grep -i netname

    NetName:        AMAZON-EC2-8

You're asking a loaded question. Companies spend money on acquisition strategies. Free trials are only a loss if they don't result in the acquisition of customers that justify the expense.

As to your question, at IndexTank we had a free trial plan that allowed us to host hundreds of customers on a single AWS instance. Why would you imagine that any company operating at scale would have to dedicate a whole instance to a nonpaying customer? It's the same as assuming that AWS would give a free user a whole physical server.

BTW, CircleCI is not bootstrapped.

http://www.crunchbase.com/organization/circle-ci

That said, when we were bootstrapped (the first 14 months or so of CircleCI's life), we still offered a free trial.

I think many companies have additional services that are free just to attract clients. For example I found a good on-line service North and Loans here http://northandloans.ca/ that provides payday loans. And this service is completely free for clients. It's a reliable service that matches the potential borrowers with the direct loan lender. So the client just need to confirm his or her solvency and fill in the application form. Everything else is done by the specialists. The money is transfered through the bank.

Typically startups just rent one or two instances big enough to handle the first set of users/customers then expand as needed. So the cost per free user depends greatly on the product. I think it is unusual to give one instance per free user.

Reply on news.ycombinator.com