Agreeable Data reposted this
I think the problem might be the pricing model? Like, when we first launched Mode, we charged different prices for creator and viewer seats. Back then, we figured that analysts and data engineers would be our power users, would value the product a lot. And we could charge a lot for those seats, and give away seats to "consumers" for free. But this doesn't really work. First, to make the economics work, you had to charge a relatively high price lot for analysts — roughly, a couple hundred dollars a month. Even if the total cost of an annual contract isn't that high, people balk at that sticker price; they're accustomed to SaaS products being tens of dollars a month, not hundreds. And second, the people who create stuff in data tools aren’t like designers or engineers, where there’s a pretty hard line between a creator and a consumer. Companies are full of people who want to occasionally write SQL queries or create their own reports. But they don’t want to do it as much as an analyst, and will argue that it’s not fair for them to pay the full analyst price for partial analyst usage. At Mode, first tried to solve this by negotiating a bunch of weird partial seat licenses, where customers would tell us they wanted 10 full seats and 50 partial seats, so we’d charge them for 20 and call it even. But this is really messy to sell and a huge pain for customers to manage. Plus, both of us were trying to triangulate around an agreeable annual price. All the partial seat antics were just gymnastics to make it work. So, we eventually threw it all out in favor of a flat per seat fee. But inevitably, this incentivized us to build more features for everyone, and one you do that, you're inevitably going to build a BI too. And so we did. Every other data product get stuck in the same trap, for what seems to be the same reason: It starts as a specialized tool; its pricing model encourages the company to build things that expand who can use the product; they build generic features like dashboards and drag-and-drop charts; they become BI. Show me your pricing page, I'll show you the outcome, I guess. https://lnkd.in/eeMDvYNU
Drew Harry
1mo
As a Mode customer -- yes. One tweak that I always wanted but could never get out of the sales people was "monthly users" pricing versus "seat" pricing. It really ground my gears to pay a roughly annual fee for the people who viewed a mode report a few times a year. It felt like maybe a third of the total cost was these incidental users who really could take it or leave it. Our reaction to this was a license-expiration regime where if you didn't log in to Mode for N months, we disable your account to save on costs. It was operationally obnoxious for both users and administrators. Plus the move into BI (which I understand is largely unavoidable) meant the rising per-seat prices were justified by capabilities we didn't use at all. A one-two punch.
William Wolf
1mo
What do you think about usage based pricing for data tools? I’ve been thinking a lot about it recently. It’s obviously secondary to finding PMF, but I think it’s odd that there aren’t many data tools that charge per number of records, per dashboard, per dashboard view. Adjacent tools in observability like Datadog and compute platforms seem to do well with these kinds of models. I’m not really sure why everyone is so fixated on per seat prices
Dexter Horthy
1mo
how about "show me the outcomes and I'll tell you the price" 😉
Jonathan March
1mo
I believe these product pricing issues can be traced back to the false presumption that data democratization is the best (or only) solution to a data team that has become a bottleneck. The root of this false presumption runs even deeper and is not an isolated problem to data teams. It is an organizational development problem (read: work culture mindset) that has destroyed efficiency and lowered the quality of work across many fields/industries.
Drew Breunig
1mo
Every data app eventually becomes BI, just as every productivity app eventually becomes a calendar. I’d say it’s the pricing model and the metrics. SaaS turns software into an annuity, emphasizing ARR, acquisition and churn, and margin. At some point, every successful SaaS co saturates their market (either by themselves or with their competitors) yet still has to grow. So more products to sell to the users they have. Should we be looking for a way out (which means new pricing models, metrics, and —necessarily—investors) or is this model sufficient?
Maria Patrakova
1mo
It gets even more complicated with embedded BI. It is a very different use case when someone checks a couple of charts twice a week from an internal user doing heavy data exploration daily. Very quickly, it becomes more cost efficient to build your own dashboards from scratch than to pay per viewer seat. I liked some ideas about usage based pricing in the comments. Another option is to provide volume based discounts rather than flat rate per seat.
Kaustav Mitra
1mo
Pricing per user and pushing folks through multiple tiers of feature gating IMO simply does not work for data tools - because as you said, the usage patterns are fairly heterogenous to have a sticker price pegged to a user, where your denominator, i.e. the user can be a multitude of things. Can you imagine the carnage, if Atlassian sold a starting tier where you get parts of Jira, parts of Confluence and parts of Bitbucket? It simply would not work.
Jonathan Hansing
1mo
(Chuckles) Haha, we're in danger. I had to stare into the abyss for a second, right after reading the paragraph on "why now." Great article, hits a little too close to home. However, I hope that the BI event horizon existed previously only because there was nowhere else to go. Once you have a baseline of differentiated functionality, the only features you can add are ones we've seen before—visualizations, filters, exporting, connectors, etc. But any new AI/BI product probably has a roadmap of years long of things that (1) were not possible even two years ago, (2) customers are constantly asking for, and (3) don't look anything like BI. Why build visualizations when users want an AI to remember and reason about business context in the form of unstructured annotations? When users want AI to construct user interfaces on the fly, depending on their workflow? When do they want AI to fetch and integrate real-world events and context into its analysis? At the risk of sounding hopelessly naive, my bet is that it's now much easier to remain stubborn about your vision. We'll see if I eat these words in a couple of years.
Doug Miller 🚀
1mo
Great post on how pricing drives customer and corporate behaviors. Unlocking pricing is tricky, but is incredibly worthwhile. As a general rule you need to charge more as more value is created, and the question for Benn Stancil is at Mode is valu created when insights are discovered, shared, and/or acted upon?
Open source pipelines - dlthub.com
1moI find per seat model to fundamentally compete against access and data democratisation and i generally hate it because to me as a team leader it's an obstacle to deploying my work instead of a facilitator. While this is probably bad business for the tool, as a data team manager I would prefer a flat rate instead of running around for approvals/credit cards. And it's really depressing when someone comes up with the idea that everyone should share 3 accounts. And if I want 200 non-technical people to be able to read and pivot a chart then this shouldn't cost me 100x over having 2 analysts. Perhaps instead charging tiered flat rates based on usage/query seconds? That would encourage people to not put 30 charts below the fold on dashboards, which is a good thing, as opposed to refusing access or sharing accounts.