Looking for feedback on my prelaunch recommendation as a service startup

Hey everyone,

I am a new member looking for feedback on a prelaunch service I am creating. In the past I have spent months building a side project before showing it to anyone - this time I thought I would try and use the lean startup idea a bit more and have just built a landing page.

The site is here: preeli.com

I am also curious how other people get feedback for a product that hasn’t launched yet - places like HN/ProductHunt seem to only accept launched products?

Link - http://preeli.com

Seems like an ambitious product - creating a generic service which uses machine learning to provide recommendations. Could you tell us a bit about your experience with machine learning? How’d you come up with the idea?

If there’s a market for this service, then there’d also be a market for a) delivering this solution as a consultant ($$$ up front) and b) delivering something along these lines as a productised server ($$ up front) before you get in to offering it as a generic SaaS ($ long SaaS ramp of death). Have you considered these options as potentially lower-risk, greater-reward way of entering and getting to know the market?

We are currently in closed alpha and will be opening up a public beta in the next few weeks. Sign up to our mailing list to be notified of when we launch (We won’t email you for any other reason).

Is this accurate? Are you weeks from launch?

What they’re advertising is pretty standard stuff. That they’re packaging it up is a good strategy.

This sounds too much for a product that isn’t even launched. Are you sure you are building the exact thing that you are advertising? I am pretty sure Machine Learning as a service has been thought a lot of times and even if you are good at it, building a generalized API or product for that would be really hard, not only in technology sense but usability and marketing sense too.

You are better off focusing on just one thing - like, ecommerce product recommendations. You can also do away with a lot of technology description. I don’t think anyone would care what database you are using to run it.

Who’s a prospect for this service? An e-commerce store? Then the description is too techy.

I agree with people above me in the thread – this project is too big, too vague and as a result won’t be fitting anyone.

Hey,

Thanks for much for the feedback. While somewhat ambitious at a base level we are really just wrapping a nice API around a standard machine learning setup. The big thing about the setup is it learns from your users actions and intents so it can easily fit a whole range of data (eCommerce, Movies, Food) without becoming overly complicated.

I have been learning ML for a couple of years now, I started off with various Coursera courses before getting serious about it. I have implemented this exact setup for two businesses now as a consultant. It usually takes me 1-2 weeks just install and configure all the software but this could fairly easily be automated and with the edition of a easy to use API would make a good product.

We do have one client using the product right now - and we could potentially launch in a few weeks but I really want to get more feedback around the idea before I open it up.

A couple of you have said the product feels to big for something that hasn’t even launched yet so I am thinking of just focusing on eCommerce to begin with. That seems like the most logical category to start with and is the area I have the most experience in. It would really be a change in the marketing more than anything but I do think it is a good idea to focus more on one area to being with.

Really interested in any other Machine Learning as a Service companies people are aware of that are aimed at SMB.

Azure and AWS have ML products but both of them require fairly indepth ML knowledge and have fairly confusing pricing models.

1 Like

The the wording on the site should talk in eCommerce language, and explain the benefits in simple phrases, without all the technical acronyms. Right now it scares off anyone who’s not a developer.

You said you have one active client - take their presentation/slides/emails and read how they describe the system to their peers. Those are the words that should be on your site.

Hey rfctr,

Yes I think you are right. I will go over the site this evening and move a lot of the technical information to another page. I will also replace a lot of the marketing information to just focus on eCommerce instead of being generic.

Thanks so much for the comments.

How’d you get this first customer? Can you find more customers the same way? Is it financially rewarding to do so? (If so: hustle. If not: charge more.) Can you ask them for referrals?

1 Like

So your customers are people with a large enough catalogue of products that they need to be able to provide specific recommendations per customer?

I presume that means you either need to sell directly to those companies (an example from my local area would be the Superdry brand) or to the companies who provide e-commerce services to large brands (i.e. your value might be someone else’s value-add to the final customer). A potential customer in that category (again from my local area) might be Anatwine who provide a platform for large brands to re-sell their products on other e-commerce sites.

How fixed are you to the “Recommendation as a Service” business model?
Have you thought what sort of exit you’d be happy with from the endeavour?
I’d have thought a product like this might be seen to offer enough competitive advantage to an e-commerce platform that they might prefer to buy it out-right than use it as a service. If you went this route you might want to investigate some of them a bit deeper. A small platform that is VC backed and hoping to grow fast might be prepared to offer you more money to get a competitive edge on larger rivals? What is the structure of this market?

How much commercially sensitive information do you require being sent to your servers to do the recommendations? Has this been raised as a concern by any of your existing customers/clients?

I do have some connections into both the companies mentioned above, let me know if you want me to attempt to pass on your product site to them (no guarantees it will get to the right person of course).

Hey Matt,

Thanks for the reply. Yeah my first client has around 10,000 products and it works really well for them. I would say you need at least 2000 products for this sort of system to add value.

At the moment I have mainly been working with smaller companies who have custom eCommerce stores and their own in house development team. Its usually a small 1-4 person team, sometimes it is custom software, sometimes it is built on top of Magento or another CMS. I wasn’t aware of companies such as Anatwine - their could definitely be a partnership possible - one were they can add our product as a value add to their customers.

We require no commercially sensitive information, it just needs user and product ID’s. If the customer wanted to keep everything 100% anonymous then they could easily hash the ID before sending it to us so we would never know the real ID. The system learns completely off users actions rather than any content.

Re: other ML as a service product, other than Azure and AWS, I’m aware of BigML.

And then quite a few small specialized products that use ML behind the scenes, but do predictive analytics for ecommerce mainly.

So I do think it’s important to focus on a high value area, that people are already doing analysis for, and it does seem like ecommerce is a place to do that, though it seems more like Enterprise customers would be more likely to have that sort of need + money to spend.

For what it’s worth, I spent a couple of years in an funded ML startup, focused around using data to improve conversions. I would say selling performance improvements is a tough business, versus selling analysis / insights, or “do it yourself” tools for analysis. We always got stuck needing to “prove” performance in individual customer scenarios, and then justify the cost vs. benefit each time.

2 Likes