WU search costs advice

Dear bubble experts
I would really appreciate some advice…

My business is a market place that sells boutique products but could easily have a million users (a rival company has 9 million). We are only in beta-launch and currently have 2000 users so far.

I have been monitoring WU usage like a hawk, and thankfully things are currently manageable - but I can see that searching is what is wracking up the units. As it’s a marketplace there is no guarantee of sales, and a user can browse and search for ages without buying anything. The thought of 1 million users doing this at the current WU costs is pretty scary. Therefore I am doing everything possible to get the search cost down.

Currently the search is performed on the page load (once logged in) and it pulls in 18 products per page. Each record has a weight of around 30kb. I can likely get this lower using a satellite method as not all fields are relevant to the initial search, so that is going to be my next step.

I am also looking into “scious search” by @zelus_pudding and “WUreducer” by @gaimed

If I understand correctly, Scious search would save WU units because an API call is cheaper than a bubble search(?) [although typesense & algolia has their own associated costs] And WUreducer could help because all my users get the same initial ‘default’ list of products - so having a cached version sent to them, rather than a unique search, sounds like a good move.

Before I embark on those paths I was just wondering if there are any other native/simple ideas to reduce WU costs when searching. Any clever tricks or ideas that i maybe haven’t considered? Ultimately Ineed this app to be as future proof an scalable as possible.

Many thanks
Dale

2 Likes

I’m curious what sort of WU you’re seeing a month with 2000 users. Purchasing a Workload Tier may help you manage costs - for example, even the base workload tier for $29/month (on top of whatever you pay to subscribe to Bubble) not only gets you an additional 200K WU, but it cuts the cost of WU overages in half from $0.30 per 1000WU to $0.15WU

1 Like

Using an external database will definitely reduce the WU usage as your searches will be api calls in front end.

1 Like

Thanks for replying msgiblin. I realise that I can upgrade and add tiers etc, and that is something that will need to happen fairly soon already. However my point is simply that I want to ensure i’m as fully optimised as possible before we start to scale. We are at an important stage where we are talking with investors etc.

Currently our usage per month is around 250k so I guess that works out around 125 WU per user. There are around 400 products listed but we are working on another ‘area’ in the site where there are 300,000 products listed - so that will increase the number of searches even further. I need to be ready for all this.

Lets say on average we’ll be expecting about 250 WU per user per month, on current trajectory… that would be 250 Million WU units required per month with 1 million users. Which on current plans would look around $20k cost per month :scream:.

Hopefully you can see my concern and why I am eager to get the consumption as low as possible. :grimacing:

Cheers

1 Like

Thanks. It seems counter-intuitive that bubble would make searching so expensive that it pushes people to use off-platform solutions or Api calls to achieve the same thing.

I think “scious search” may be the best option because it would store a ‘lightweight clone’ of our database and return results via API. This feels like the best of both worlds. Also Andrew seems like a top guy and very helpful and looks like the product will get full support.

My only concern there is that typesense comes with its own set of costs. So will need to weigh those up too…

Okay, that’s about $0.04 per user per month. Surely that’s financially viable? You’ll be paying dollars to acquire a single user from advertising. Or, if you have 500,000 users (which let’s be honest you will probably not, at least not in the next year or two), surely you’d have enough revenue to justify building it out on a traditional codebase if it was deemed necessary?

How did you get 0.04?

Currently we have 2000 users and it’s using 250,000 WU per month. 90% of this is on the searches. So that is roughly 125 WU per user per month. When we add the additional search areas, this will likely increase further so I picked 250 WU per user (per month) as an estimate.

250 WU per user per month x 500,000 users = 125 MILLION WU units required., per month. [250 Million required when we reach 1 million users].

?

In the first place I don’t know why to start an e-commerce in Bubble and build all from scratch if there is several other options already built for this like WooCommerce and Shopify (and much more cheaper as well).

Seams a lot of work to reinvent the wheel…

@daleaskew why did you decide to use Bubble for this!?

1 Like

Don’t get ahead of yourself…

If you’re doing a large amount of WU (more than 40 million) Bubble looks into custom pricing or Enterprise plan (which doesn’t count WU as it’s just an amount of computing power).

I promise you, thinking about what you’re going to do ‘when’ you reach 1 million users is premature. Average marketplace CAGR is 16%. If you get there you’ll have enough money to decide what you want to do.

Good to know about the Enterprise plan having its own set of rules. That gives some peace of mind for sure.

Regarding getting ahead of myself etc… I built this as a small venture, totally bootstrapped and had slow growth in mind, but things have now got a bit crazy with investors and incredible partners onboard who, at the touch of a button, could get us 1 million users overnight due to their position in the industry. This is amazing and exciting but also obviously scary and why I am thinking of all these wu related scenarios and it’s my job to ensure our platform can handle it. I’ve literally been asked to do a presentation based on what ‘one million users’ will look like, including costs etc. so that is where things currently are, and why I am asking these questions…

1 Like

I’m sure some would say that. But, remember, they want to take as much of your company as possible for as little as possible, and will say what they need to in order to get that, so treat it with a healthy dose of skepticism.

Talk to Bubble Enterprise team about dedicated costs (though they might be reluctant to give exact numbers if you’re not getting it immediately) and that’ll be a good starting point.

Bear in mind even $20K a month of WU is the same price as just one senior software engineer…

1 Like

Because it was and still is the best fit for my needs, budget and skillset.

The needs of the platform goes way beyond anything that shopify or other such platforms offer. The level of tailoring behind the scenes is very deep and unique. For example our platform generates fully legal music agreements including publishing and masters, on the fly and automatically customised to specific user selected options. There’s deep bidding features, search filters , credits system, holds system and lots more. The backend admin is equally deep with scheduled mailouts, housekeeping, stats updates and much more. The interface and user experience has been extremely well thought out and implemented. Also, artists themselves are able to upload their products, which then goes through a review process, file optimisations, and other checks, again all handled via the platform and very specific/neich.

What i have been able to accomplish using bubble, as a solo developer, in 1 year is incredible and I am so grateful it exists.

A really quick overview of our platform can be seen here: https://www.youtube.com/watch?v=TQ6JT7eH79c

Hope you like it.

1 Like

For 2000 users your units should not sky rocket. Here’s my checklist you can give it a look

  1. Check if there are too many searches on page load
  2. Create reusable elements for repeating groups for them to load separately
  3. Create a separate page for products page where people can only see the list of products available all other interations should be on a separate page (so which ever page users will only scroll on should be completely separate and should not have too many searches on them)

Hope this helps

2 Likes

How many products do you all have in your Bubble apps? Mine has around 10,000, but the monthly WU cost is crazy - easily over half a million. It’s frustrating because most of that WU usage is just for displaying text!

While Bubble is still great for many things, the WU system has definitely caused some stress. Algolia and Scious Search seem like promising alternatives to reduce search-related costs. We’ve been using Algolia a bit and are considering a full switch to scious or algolia, which could potentially cut our WU costs in half.

Another major pain point has been managing updates and changes with complex recursive workflows. For years, I’ve been hoping for a better backend system in Bubble, and it’s even a highly requested feature. They promised they would get to with the introduction of WU units and I still haven’t heard or seen anything I get my hopes up every monthly update.

For B2C marketplaces on Bubble, it’s definitely possible, but I recommend exploring external tools early on. Xano is a popular option for backend management, and we’ve been testing Parabola.io as a serverless backend with some interesting AI features that have been a big time saver.

Also the use of satellites used to be great on the old pricing system. But now I need to update 20,000 products price inventory and status instead of just 10 thousand. So theres probably another quarter million wu units a month as the features need to run daily.

Overall bubble is still great for B2C don’t let anyone tell you to just use shopify or Woocommerce we wouldn’t have nearly the same amount of success if we hadn’t Choosen bubble.io. The ability to build and test potential necessary features is amazing.

TLDR: If you are going to be a B2C you can’t just rely on bubble platform. Scious, Algolia, WuReducer, Xano Parabola are all things you should look into early on.

I’m still currently on the old $32 a month legacy

1 Like

Thank you Mack it seems you are in a similar place but a bit further along the road. Currently have only got around 400 products live but we have thousands more on the way.

Yes looking at those optimisations now rather than later is what I’m attempting to do. Better to address it now than later, right :slight_smile: It’s actually kinda quite a fun challenge trying to squeeze out as much efficiency as possible.

I hadn’t heard of Parabola so thanks for that, will investigate.

Best

Rebuild in Flutterflow if you require capacity/scale/bulk costings once the MVP bit is done. Bubble is good for prototyping at the expense of paying users and then moving away once you’re growing.

I looked into flutterflow at one point, but got turned off by some things. I think one of them was scheduled backend workflows, which I use a lot. Seemed a bit of a ballache in flutterflow.

To date bubble seems to be the best all-rounder, just a shame that search wu cost is so high…

anyway will be implementing scious and wureducer soon and will report back with real world findings.

it seems counterintuitive to make search at scale more expensive on bubble than outside, but unfortunately, if you check my replies in the forum with most upvotes, thats what they did.

i just moved another client to xano to reduce costs massively.

2 Likes