Benefit Layers: How to avoid sales fluff in devtools
dx.tips
dx.tips
Show me your drawbacks and I might consider you are a serious vendor.
Can’t remember any other product from any market openly showing what it can’t do, why are devtools any different?
most founders refuse to admit that they're not fit for purpose for some things. yes need to promise the world to VCs but also it makes it harder for the developer to assess how much to trust you if you are in denial about TAM
I'll only get to drawbacks/limitations once I get to documentation. And once I spend a significant amount of time researching a product. And when I don't I feel like I'm buying the marketing.
ChatGPT, ironically enough given how many times I see people insisting that them saying “we can be dangerously wrong don’t rely on us” is some kind of marketing scheme.
One thing I've consistently had to relearn in developer marketing:
"Talk benefits, not features" doesn't work!
Developers are sick of vague promises and wary of black boxes.
Instead:
- Demo - reach Wow! in <10 mins
- Explain how it works
- Show how real companies use in prod
and that really resonated with folks (heard some really good feedback from senior github people). excited to return to this topic with a better framework (https://pbs.twimg.com/media/F03sDA6agAEBf_M?format=png&name=...) for when to communicate with benefits vs features
It obviously doesn't work.
Can you imagine a restaurant flyer that just says "Enjoy delicious meals! Have fun time!"
It's absurd.
You show pictures of the food. Tell people how you source your ingredients. How your chefs have perfected their craftsmanship. Etc etc.
It only appears to work if you think that kind of advertising is how they got to their market position. Usually it's not.
Let's Docker for an example.
Docker did not get adoption because of some cliche marketing on their landing page. It got adoption from people doing tech talks and tech demonstrations and several companies writing blog posts about how migrating to Docker saved them from the headaches of deployment they were having before with the ad-hoc systems.
Complex enterprise tech gets sold by other means, usually personal relationships with key decision makers, and maybe some amount of bribery, who knows.
Docker failed to sell its commercial products to enterprises relative to the OSS adoption of its container runtime, at least at the venture scales they committed themselves to
Afaict what was being sold:
* Artifactory (hub): plenty of companies doing fine.. for product different than what docker did
* DC OS (swarm): managed k8s etc doing fine too.. but diff product again
They seemed close to PMF for both... But never quite hit it?
Selling to enterprises is hard.. I'm not sure what the lesson is here beyond git stars not being revenue, nor guaranteeing that it can turn into revenue. Successfully building one thing but selling another means you have to do 2 hard things not just 1.
Maybe a good analogy: Bloomberg has great reporters whose content goes out for free... But that's not what sells the terminal and data feeds
Benefits to pique their interest and draw them in. Features to tick the checkboxes and get through procurement.
Same for cigarettes and other similar products.
Some people may prefer the flavor of one product over the other, but at the end, it's purely about lifestyle and the feeling of a brand, and that is shaped by marketing.
100% anyone who tries this will fail
People go to restaurants for the dining experience.
If they simply wanted good food they'd eat at home that night. If they simply don't want to cook they'd get fast food.
Declining both those options means that they want something more than food which is provided by the restaurant.
The other thing that I want to know, where I’ve had remorse in the past, is how much effort it’s going to be to integrate the tool with my current processes and workflows. I don’t want to change my processes, and I don’t want to find that I cannot get that wow because I’m using this homebuilt process over here or this specific third party tool over there or because we’re migrating to cloud in 3 months.
I don't think this is true... It's more nuanced than this. Developers are much more skeptical of a "benefits" pitch, and care more about a "features" (and limitations as mentioned above) pitch. However, if you're just pitching a developer, you're probably only pitching the user, not the buyer. The buyer VERY much cares about the benefits, otherwise they're not buying. The buyer won't using a single feature, so that type of marketing is lost on them.
This is the challenge: the right message (features or benefits) to the right person (user or buyer) AND at the right time!
IOW, the pain you think they are feeling is not the pain they are actually feeling.
As someone also trying to bootstrap a product, I feel your pain, but it's not specific to developers.
I know personally that waiting 15 minutes for a test suite to finish is super frustrating, but people seem to have normalized it. I've also literally been hired to speed up people's tests, so it's not a problem that doesn't exist. Our current customers are also delighted with the speedups so I know it's a problem. But explaining that problem and why the status quo is bad to developers is especially hard. Btw I don't buy the "developers are immune to advertising" spiel, I actually created some of CircleCIs first ads so I have experience doing this. I just think developers are different and marketing devtools is different again (because of it's not a personal purchase and there are a lot of different stakeholders).
They probably don't notice that delay, because they are doing something else during that time ... Reading the next ticket, some git admin, replying to an email, checking slack, etc.
As a single example, I don't consider a 15m delay for a full test suite to be a problem. I've got plenty of 5m tasks to do that I can use to fill in that time.
I cannot think of any developer who sits and watches the tests for execute.
I think some developers don't care, and some developers really really care and I haven't found a good way of partitioning the set and focusing on the developers who care.
You need to find the target market where a 15m delay is a pain point i.e. partitioning the set.
I don't personally know of any Devs who consider that a pain point, so it might be a really small target.
Trust me, I am going through the same thing, and I'm constantly questioning whether this yet to be released product has a big enough target market. No good to me if the alpha users comprise the entire target market, so I'm ready to pivot the moment they say "this is great, but we don't have budget/use this less great free thing/can't get approval/will sign up next year/etc".
That's all code for "this isn't a pain point for us."
The uncertain "faster build time" would be nice. But is it worth raising it up and battling for it? Unlikely.
I think a different case can be made if you target the managers instead. "Make developers twice as productive" might be a better selling point.
Compliance, reporting, scaling with number of projects/growth, interop with existing tools, and lastly a good word from other people in the same industry are what's going go sell your product.
In enterprise CI there are a bunch of other tools you're going to be waiting on, and at least from what I've seen on the field companies will run stuff like SCA/SBOM every build and keep a history graph of when and whom added things. The developer themselves typically does not care for that, but the 'enterprise' does
And that's in big part because they have been burned before. Someone bought a Product[tm] that supposedly Solved[tm] a Problem[tm]. Which then failed miserably to do anything of the sort. Then, to add insult to injury, these same developers are burdened with the need to maintain the expensive and useless thing that they can't get rid of.
If you are trying to sell a devtool, it should have next to zero cognitive burden even in the pathological case. Moreover, you absolutely can not lie, at all, about what your tool does or does not do.
Oh, and once the tool has proven to be useless, ripping it out of the stack should be friction-free. Easy? Hell no. But necessary? Yes.
Turned off vimium and ublock and still the case. Don't think I have any weird js going on. Also happens in incognito. I'm on desktop.
I believe that in certain scenarios it will be much faster. But I would have to go evaluate if it works in my specific case and how much effort it would take to move from our current setup to a new platform.
Combine that with the fact that 15 minutes of tests really isn't that bad. I came from the semiconductor industry, where some simulations would take days/weeks, that was a really frustrating feedback loop.
The 15 minutes is only really frustrating in one case, where you are working on one high priority feature/bug fix that needs to be fixed and deployed ASAP. Those 15 minutes can be killer when you've got some bug crashing your server.
Maybe you could monitor incident post-mortems and attempt to sell to those companies. Chances are, in the incident report the time it took for the tests to run will be highlighted.
The bit in your welcome blog about local tests being unfeasible is interesting. Maybe there's a pain point around pushing untested code? Like wasting reviewers time.
Drop me a line if you want to chat it through, sounds like an interesting challenge.
It might be the case that we think we want no fluff, but we're actually swayed more by fluff.
you still need principles to drive at the end of the day
1. Dev visits the site, gets convinced that he wants the tool.
2. Dev convinced management to allow the expense.
3. Dev visits the site and buys the tool.
You want to test step 1 but only get data for step 3, where the content of the site is completely irrelevant.
Got me a first round at Google. Unfortunately for them, I super suck and therefore failed in the first round. Ain't nobody got time for leet code when they already have a job.
In my defense, I was more respectful than that. Also, the "we should use real technology" was my response at Time Magazine, not at Google. I stand by it. Server-side javascript (isomorphic javascript) is stupid. At least try to use typescript or something. The interviewer clearly agreed with me. Problem is it was above their pay grade as well.