Revenue-Weighted Prioritization: How Savio Ranks Feature Requests by MRR

Five votes from $50 customers, or one vote from a $5,000 customer. A vote counter says the first request wins by five to one. Weighted by revenue it is $250 against $5,000, and the second request wins. Savio built its whole product around that second answer, so we spent an evening inside a trial workspace to study how they did it.

This is a Field Notes piece: an evening inside the product, the parts we admire, the rough edges we hit, and the version we built for Usero before the night was out. The Savio screenshots come from a hands-on 14-day trial with sample data, no card, no sales call.

An attribute, not a field

The first design decision is the one we kept coming back to. Savio has no hardcoded “MRR” field. Instead, any company attribute can be flagged as “stores the customer’s MRR” and gets a dollar icon. The default attribute is named monthly_spend, the display name is yours to set (the sample workspace calls it “Sample MRR”), and the flag is what lights up every revenue surface in the app.

The payoff of doing it this way: one attribute system powers columns, filters, contact panels, and report metrics for anything a company record holds. Plan, seat count, renewal date, MRR, all the same machinery. The MRR flag is a single bit of metadata on top. We would guess this fell out of building the attribute system first and the revenue feature second.

The join that makes it trustworthy sits one layer down. Every piece of feedback in Savio requires a contact, and contacts belong to companies. That discipline is enforced at the create-feedback form, whose first field is “The problem this customer has is (include verbatim if possible)”. Every dollar in a request’s total traces back to one contact’s feedback, so a request with no contact attached has no number at all.

Who is asking, as dollars

Once the attribute is flagged, the feature request list can show it as a column. The number per row is the sum of MRR across all companies whose feedback links to that request, and the column sorts. Sorting by it is the core view of the product: your backlog ranked by how much monthly revenue is attached to each ask.

A small wrinkle worth noting: the MRR column is not built in. You add it yourself via “Pick attribute to display on FR list” in the column settings, and a banner walks you through connecting customer data first. More on that trek later.

Open a request and the sum decomposes. The sidebar shows the cumulative figure next to counts of requests, companies, and feedback items, and each linked piece of feedback carries a per-company chip (“Sample MRR: 1,500”) with an inline pencil to edit it. You can see exactly which customers make up the total, which keeps the big number honest.

The manual fallback is a feature

Here is a choice we want to applaud. The Edit Company form is three things: a name, a Sample MRR box, and an “Advanced: Show Sync Info” toggle. The whole revenue layer runs on hand-entered numbers, no integration required. That is how our trial workspace worked for the entire evening, and it is why the feature demos on day one instead of after a week of CRM plumbing.

The sharpest idea: filters recompute the dollars

This is the detail we came away wanting to steal. Filter the list by the MRR attribute (“Sample MRR >= 1000”) and Savio recomputes the sums instead of hiding rows. Take one row from our workspace. The Zapier Integration request totaled 3,750 MRR unfiltered, summed across 3 contributing entries. Apply the “Sample MRR >= 1000” filter and the one entry from a company paying under $1,000 drops out of the sum, not just out of view, so the same row recomputes to 3,000 across 2 and the list re-ranks around the new total. Now the column measures revenue only from companies over your threshold. A PM segmenting by spend wants exactly that, and most tools would have shipped show/hide and called it done. The catch: the row keeps its “Sample MRR” label while the number underneath changes meaning. The 3,000 you see filtered is not the 3,750 from a minute earlier, so it is easy to carry a filtered figure in your head and compare it against an unfiltered one later.

The same idea carries into reports. “Top Feature Requests by Time Period” has a Metric dropdown with Total Feedback, Unique Requests, Unique Companies, and Sample MRR, and picking the last one turns the chart into a revenue-weighted view. The option’s form value is a dynamic attribute id (fa_61474 in our workspace), so reports run off the flagged attribute too, the same machinery end to end.

The rough edges we hit

A few honest notes from the same evening, offered in the spirit of a team that has shipped plenty of rough edges of its own.

• The filter and the column measure different things. The column is a cumulative sum across companies; the filter tests each company individually. We set “Sample MRR >= 3700” expecting to catch the 3,750 row and got zero results, because no single company on that request pays 3,700. Correct behavior, and still a trap. A word of copy near the filter would save the next person our five minutes of confusion.
• Setup is a trek. The full loop wants a support tool for feedback, a customer source of truth for attributes, an attribute flagged, and a column configured, spread across several settings pages. A “Finish Setting up Savio (7 left)” checklist follows you the whole way, which is a fair acknowledgment of how many steps there are.
• The automated path costs $249 a month. Salesforce, HubSpot, and Segment sync live on the Business tier. Essential at $39 and Professional at $79 get revenue weighting through support-tool attributes or manual edits. We should flag that we did not connect a real source of truth ourselves: the auto-import and its nightly-or-faster sync cadence are what Savio’s docs describe, not something we observed running.
• No AI in the loop. Linking feedback to feature requests, deduping, and grouping are all manual in Savio. The revenue math is only as good as the triage feeding it, and the triage is a human reading tickets.

None of that dents the core design. The attribute system, the mandatory contact link, and the re-aggregating filters are careful work, and studying them shaped what we built next.

What we built the same night

Usero already had the plumbing Savio’s design depends on: a Person record per user, feedback linked to people, and an identify() call that accepts arbitrary traits. What was missing was the revenue layer on top. So we added it.

Pass identify({ traits: { mrr: 499 } }) from your own code, one line next to wherever you already identify the user at login, and the number lands on the person. The inbox and the AI clusters then show revenue sums and can sort by them. Our ICP is developers shipping their own product, and their code already knows what each customer pays, so the number can come from the source instead of routing through a CRM. That is the piece Savio gates at $249 a month, and for a developer it is an afternoon of nothing.

Two choices we made deliberately after the evening in Savio. Sums are computed over distinct people, so five tickets from one $500 customer stay $500 rather than inflating to $2,500. And the weight lands on clusters, the groups our AI already builds from raw feedback, so “$X of MRR is asking for this theme” appears without anyone manually curating feature requests first. Savio weights the feature requests a human triaged into existence. We wanted the dollar figure to land on the groups that form on their own.

Scope honesty: this is a first slice, sitting in review on a branch as we publish, not yet live for customers. No CSV import of revenue yet, no Stripe sync, no revenue-weighted reports. Savio’s re-aggregating filters are on our list too, because that idea is too good to leave in their app.

The takeaway

Savio’s revenue weighting is a well-built answer to a real question, and the craft is in the quiet decisions: an attribute flag instead of a hardcoded field, a mandatory contact on every piece of feedback, filters that recompute the dollars. The cost is setup work and a $249 tier for the automated feed, and the numbers rank whatever a human has manually triaged. We took the ideas we admired, attached them to clustering that runs on its own, and fed the revenue in through one line of code.

If teardowns like this are your thing, the first Field Notes piece walked through how Canny Autopilot works. Or you can try Usero free and see what the clustering finds in your own feedback.

Frequently Asked Questions