[HN Gopher] Launch HN: Promi (YC S24) - Personalize e-commerce d...
       ___________________________________________________________________
        
       Launch HN: Promi (YC S24) - Personalize e-commerce discounts and
       retail offers
        
       Hey HN! I'm Peter from Promi (https://www.promi.ai/). We're
       building a platform for ecommerce merchants to send realtime
       personalized discounts, optimized with AI (obviously)  Demo:
       https://youtu.be/BCYNCqb4fUc, Sales Video:
       https://www.youtube.com/watch?v=WiO1S7RBn-o  All the big tech
       companies send personalized discounts - Uber, DoorDash, Google,
       etc. In fact, I was the product lead overseeing discounts at Uber,
       so if you've gotten a promotion on Uber Rides or Eats, that was our
       tech. These personalization models often generate >30% more revenue
       vs. non-personalized discounts (cost-neutral that is), so this is a
       hugely impactful product.  It's no surprise then that other
       merchants want to follow suit. Merchants don't want to waste
       discounts on customers who would have purchased anyway. Frankly
       it's not a new idea to offer a software solution to personalize
       discounts - plenty of other startups have entered this space with a
       similar product.  The biggest problem with personalizing discounts
       for mid-size and smaller companies has been that traditionally you
       rely on 'explore' data - data from randomly sending out discounts
       to a portion of the user base. But this has a lot of problems:
       merchants need to be large, collecting this data is expensive,
       training data really should be fresh (so explores should constantly
       be running), and if you want to try a different discount structure
       (e.g. BOGO instead of 20% off) you'll need to run a new explore
       with the new structure.  So what does Promi do differently? We
       train on regular traffic and simplify the problem by just focusing
       on conversion rate. If we can accurately predict who is unlikely to
       convert and which products are unlikely to be bought, we can issue
       discounts without the fear of burning money on an order that would
       have happened anyway. One of my major takeaways from my time at
       Uber was that our model was mostly targeting users who had a low
       likelihood of converting in a given week. Quantifying how much more
       likely they were to convert when given a discount via explores was
       helpful, but not as impactful as understanding starting conversion
       rate.  Side note - It's been a bit interesting launching an AI
       company during this hype cycle that isn't actually using the latest
       and greatest LLMs. We believe more traditional machine learning
       still has a lot of value to add. I don't want to say we won't use
       LLMs down the road (there may be some interesting applications for
       developing additional features), but starting this way has worked
       out well for us.  There have been plenty of other challenges (as
       with any startup). We've had to figure out how to automate
       integrations when so many websites have custom code. We've had to
       make the model work without rich user data since the majority of
       website visitors aren't logged in. A quick note in this one - we
       can use first party cookies to more or less track the view and
       transaction history, but we've found that one big predictor of
       conversion is traffic source: whether a visitor is coming from ads,
       email, direct traffic, google search, etc. That traffic source
       isn't something as valuable at Uber (since everyone uses the app),
       so it's been a bit of a tradeoff in the types of features that are
       most impactful.  Our model seems to be working well! We have case
       studies on our website showing the typical revenue and profit lift
       we see. We currently have tiered pricing with different quotas for
       the amount of revenue managed by Promi discounts.  I'd love to get
       thoughts from the machine learning experts in this community,
       though full disclosure I'm the non-technical founder. Let us know
       what you think!
        
       Author : pmoot
       Score  : 16 points
       Date   : 2025-07-22 16:11 UTC (6 hours ago)
        
       | lazyninja987 wrote:
       | Does a merchant has to give your tool access to their user data
       | to generate personalized discounts? Apart from user activity
       | data, what data do you need for maximum effectiveness?
        
         | pmoot wrote:
         | Yes. We're going through Shopify, so merchants agree to terms
         | when they install the app.
         | 
         | There's user activity data, but also contextual data and shop
         | data that we use. 'Contextual' data refers to things like
         | device type, traffic source, time of day, day of week (there
         | have been some interesting trends with corporate vs. non-
         | corporate customers in this one).
         | 
         | Shop data includes things like product profit margin and
         | product conversion rate. Obviously we can go deeper with our
         | discounts on products that are very profitable, and it's
         | typically more efficient to give a discount on products with
         | lower conversion. Merchants also like boosting items that
         | haven't been selling well.
        
       | klaussilveira wrote:
       | > If we can accurately predict who is unlikely to convert
       | 
       | Do you use historical purchase data to make that assumption? Or
       | someone that frequently abandons carts?
        
         | pmoot wrote:
         | We use historical purchase data, as well as view history,
         | traffic source, device type, etc.
         | 
         | Traffic source a lot of times is the most impactful. People
         | coming from ads are often more in a browsing mindset, vs.
         | people typing in the url directly have a higher purchase
         | intent.
         | 
         | We don't have abandoned cart rate as a feature in our model,
         | but actually might be something worth looking into adding.
        
       | malshe wrote:
       | If I understand it correctly, you estimate the probability of
       | purchase given the user characteristics, behaviors, etc. If this
       | probability is below a cutoff, you offer a discount. Did I get it
       | right?
       | 
       | Is the cutoff itself a function of other variables in the data?
        
         | pmoot wrote:
         | Yes, that's mostly right. We also vary the discount value, so
         | it's less a binary discount/no discount and more a range. There
         | is often a cutoff though. Merchants can input a hard cutoff
         | e.g. if they want to ensure everyone gets a discount (great if
         | they also have marketing assets for a sale), or if they want to
         | avoid making their sites feel too 'sales-y'. Otherwise the
         | cutoff is defined by conversion prediction, inventory levels,
         | and a few other inputs.
         | 
         | There's actually a lot more we could do to make this cutoff
         | more intelligent though - e.g. at Uber the cutoff was set to
         | exhaust a certain promotional budget. Or we could target a
         | specific ROI if we eventually have good enough predictions.
        
           | malshe wrote:
           | Thanks for the reply. Do you use Bayesian models for this?
           | Btw, Pete Fader[1] has done so much work in customer
           | valuation where estimating the probability of purchase is a
           | crucial aspect. Maybe you already use them.
           | 
           | [1]
           | https://marketing.wharton.upenn.edu/profile/faderp/#overview
        
       ___________________________________________________________________
       (page generated 2025-07-22 23:00 UTC)