Deck

How-To Guide12 min read

Why Feature Voting Fails, and How to Rank Requests by ARR and Churn Risk Instead

Last updated: · By Matt Teixeira, Founder and CEO

On this page
Short answer

To rank feature requests by ARR and churn risk instead of votes, score the accounts behind each request, not the votes on it. Count each paying account once, add its ARR up to a cap, double the weight of any at-risk account that names the request as a reason, and discount accounts outside your target segment. Then adjust for how many sources raised the request and how well it fits this quarter's strategy. Votes stay in the model as one source of evidence. They just stop being the ranking.

Key Takeaways

  • Votes measure who visits your board, not who pays you or who is about to leave.
  • Count each paying account once per request, however many of its users voted.
  • Cap each account's ARR near your 90th-percentile contract, so one big customer can lift a request but not decide it.
  • Apply churn risk only when the account names the request in a renewal call, cancellation reason, or escalation.
  • Give each free account an expected value: median paid ARR times your free-to-paid conversion rate.
  • Keep your board for intake and for telling people what shipped. Rank from the score.

Why Does Feature Voting Fail?

A voting board measures enthusiasm among the people who visit it. That group is small, leans toward free and self-serve users, and leaves out your quietest, most valuable accounts. Each failure has a fix in the model below.

FailureWhat it looks likeFix in the score
A loud minorityA few power users post often and rally their colleaguesCount unique accounts, not votes
Free users outvote paying ones140 free workspaces outvote 7 enterprise accountsWeight by ARR; give free accounts an expected value
Silent signals go missingEnterprise admins tell their CSM; churning customers cancel without postingPull from calls, tickets, and CRM notes too
Duplicates split the vote"Bulk edit," "update many rows," and "mass change owner" sit in three postsMerge by problem before scoring
Votes never expireOld posts keep their totals after the voters churnUse a rolling window and current ARR
A vote costs nothing"Nice to have" and "we will leave without it" both count as oneFlag churn risk tied to the request

The pattern is well known. The "1% rule" of online communities holds that only about 1% of users create content (Wikipedia), and Metabase notes that "lifetime totals always favor old requests" (Metabase).

The deeper problem is what votes reward. Easy-to-like ideas such as dark mode collect votes from anyone, while a request that blocks an enterprise renewal may collect none. One sentence from that renewal call should outweigh 200 upvotes.

When Is a Voting Board Still Worth Having?

Keep a board when your users are your buyers, when customers need somewhere to send ideas, or when you want to tell them something shipped. Just don't let its vote count rank your roadmap.

  • Self-serve products: the person voting is often the person paying, so votes track revenue better than in sales-led B2B.
  • Long-tail ideas and context: boards catch small ideas that never come up on a call, and the comments often beat the count as evidence.
  • Closing the loop: every voter is a ready-made list to notify. Canny, for example, pairs boards with a public roadmap and changelog (Canny pricing). The close-the-loop guide covers the requesters a board never captured.

Boards have gotten smarter, too. Canny scores can include MRR and count each company once (Canny Help). But revenue weighting only sees who reached the board, so treat the board as one source among several. If you're thinking about replacing yours, see Canny alternatives.

What Goes into a Revenue-Weighted Request Score?

The score has two levels: a weight for each account behind a request, and two multipliers for the request as a whole. Every value below is a starting rule of thumb, not research.

InputApplies toHow to measureStarting values
ARREach accountCurrent ARR from your CRM or billing systemCap at your 90th-percentile account
Segment fitEach accountDoes the account match the customers you're trying to win?1.0 core, 0.6 adjacent, 0.3 outside
Churn riskEach accountAt risk, and names this request as a reason2.0 if tied, otherwise 1.0
Source breadthEach requestDistinct source types: board, tickets, calls, interviews, CRM notes, NPS0.8 for one, 1.0 for two, 1.2 for three or more
Strategic fitEach requestDoes it move this quarter's outcome?0.5 off strategy, 1.0 neutral, 1.5 direct
Markdown template
Account weight      = min(ARR, cap) × segment fit × churn multiplier
Free account weight = median paid ARR × free-to-paid conversion rate
Request score       = sum of account weights × source breadth × strategic fit

Read the score as weighted ARR, not real dollars: how much of the revenue you want to keep and grow sits behind this problem? The multipliers are deliberate. In a sum of 1-to-10 ratings, a guessed "strategic fit: 9" can outweigh real revenue. Here, strategy can raise a score by half or cut it in half, and no further.

Step 1: Merge Duplicate Requests into One Problem

Before you count anything, merge posts, tickets, and call notes that describe the same problem. "Bulk edit," "update 50 rows at once," and "mass change owner" become one row: changing many records at once takes too long. Keep a link to every original item, because the people behind them are your notify list later.

Group on the problem, not the proposed fix, so two solutions to one pain don't split its score. For the full grouping method, see how to decide what to build next.

Step 2: Map Every Request to an Account

Votes come from users, but revenue comes from accounts. Join each voter, ticket requester, and call participant to a company record by email domain or CRM ID, then count each account once per request.

  • Use a rolling window: the last 180 days, or 90 if you get more than a few hundred requests a month.
  • Drop dead weight: churned accounts and expired trials leave the ARR column.
  • Watch the ratio: three or more votes per account means a few teams are rallying colleagues. That's enthusiasm, not reach.

Step 3: Add Capped ARR and Segment Fit

Cap each account's current ARR at about your 90th-percentile account, then multiply by segment fit. Without a cap, one $400,000 customer equals 33 customers at $12,000. With a $60,000 cap, it still counts five times as much as a median account, which is usually the right amount of pull.

Productboard's Customer Importance Score uses the same idea, capping any one company at three points per item (Productboard).

Write your segment definitions before you score, or segment fit becomes a lever for whichever request someone already likes. The define segments playbook covers tagging feedback by plan, company size, and key account.

Step 4: Flag Churn Risk Only When the Request Is the Reason

An account that is at risk and asks for something is not the same as an account at risk because of it. Apply the 2.0 multiplier only when both are true:

  1. The account is at risk: a red health score, a downgrade, or a renewal in the next two quarters with open concerns.
  2. The account names this request, or the problem behind it, in a renewal call, cancellation survey, escalation, or CSM health note.

Have your customer success lead confirm the flags monthly. It usually takes about 15 minutes and keeps "at risk" from becoming a label everyone stretches. Show open pipeline beside the score, not in it, since prospects ask for plenty that never closes.

Step 5: Adjust for Source Breadth and Strategic Fit

A problem heard in tickets, sales calls, and interviews is more likely to be real and widespread than one heard 50 times in one channel. Count distinct source types, with the board as one of them, and apply the breadth multiplier.

Set strategic fit once per quarter, for every request, in one meeting, using only the three values. If someone wants to change one request's fit mid-quarter, the real conversation is about changing the strategy.

Step 6: Rank, Stress-Test the Top Spot, Then Sequence by Effort

Sort by score, then try to break the top spot. Nudge each weight one step, such as the churn multiplier from 2.0 to 1.5. If the #1 changes, the call is close: talk it through instead of letting a decimal decide.

Bring in effort last. Divide each score by person-weeks to set the build order, so cheap fixes change the sequence but never crowd important problems off the shortlist. Re-run the score monthly and reset the weights quarterly.

Worked Example: Four Requests Ranked by Votes and by Score

These numbers are illustrative, not real customer data. A B2B SaaS company has $6 million in ARR across 300 paying accounts, plus free workspaces. Median paid ARR is $12,000, the cap is $60,000, and 4% of free workspaces convert, so each free account counts as $480. The quarter's goal is net revenue retention in core accounts.

RequestVotesAccounts behind itPaying ARR (raw)SourcesStrategic fit
A. Bulk edit records18222 paying (16 SMB, 6 core), 140 free$180,000Board, tickets1.0
B. Audit log export97 core enterprise; 2 at risk and citing it$520,000Sales calls, renewal notes, tickets1.5
C. Approval workflows3418 paying (14 core, 4 SMB); 1 at risk and citing it$310,000Board, interviews, tickets1.0
D. Mobile app1208 paying SMB, 96 free$40,000Board0.5

Applying the formula, with scores in thousands:

RequestSum of account weights× Breadth× StrategyScoreRank by votesRank by score
A. Bulk edit records$208,8001.01.020913
B. Audit log export$445,0001.21.580141
C. Approval workflows$309,0001.21.037132
D. Mobile app$70,0800.80.52824
  • B went from last to first. The cap cut its $520,000 to $340,000, and two accounts that named audit logs in renewal calls added $105,000. Those admins talk to their CSM, not the board.
  • D fell hardest. Most of its votes came from 96 free workspaces, it showed up in one source, and mobile is off strategy this year.
  • The top spot holds. At neutral strategic fit, B still scores 534; with a 1.5 churn multiplier, 707. Even at direct fit, C reaches only 556.
  • Effort set the order. At 4, 2, and 8 person-weeks, B returns 200 points a week, A 104, and C 46, so the team ships B, then A, then starts C. The board was right that bulk edit matters, just wrong about how much.

Common Mistakes to Avoid

  • Treating the score as the decision. It ranks the evidence. Strategy, timing, and dependencies still need a person.
  • Moving weights to rescue a favorite. Publish the weights at the start of the quarter, and change a score only by adding evidence.
  • Forgetting the voters. A request that ranks low still deserves a reply saying where it stands.

Tools That Help

A spreadsheet runs this model well up to a few hundred requests; tools save the account matching and keep quotes attached. For a ranked comparison, see the best tools to prioritize feature requests by revenue.

  • Spreadsheet plus a CRM export (free): one row per account and request, then a pivot by request. Template below.
  • Canny: best if you want to keep a public board and add revenue to it. Scores divide weighted impact, such as MRR and opportunity revenue, by effort (Canny Help). Pro starts at $79 a month billed yearly; the HubSpot and Salesforce integrations need the custom-priced Business plan (Canny pricing).
Markdown template
| Request (problem) | Account | ARR | Capped ARR | Segment fit | Churn multiplier | Account weight | Source type |
|---|---|---|---|---|---|---|---|
| [Problem statement] | [Account] | [$] | =MIN(ARR, cap) | [1.0 / 0.6 / 0.3] | [1.0 or 2.0] | =Capped ARR × Segment fit × Churn multiplier | [Board / Tickets / Calls / Interviews / CRM notes / NPS] |

Pivot by request: score = SUM(Account weight) × source breadth (0.8 / 1.0 / 1.2) × strategic fit (0.5 / 1.0 / 1.5)
Show beside it: votes, unique accounts, raw ARR, open pipeline, effort in person-weeks

Where Deck Fits

Deck automates most of Steps 1 and 2, plus much of segment fit and source breadth. The revenue math stays with you.

  • Merging and accounts: Deck pulls feedback from support tools, Slack, calls, surveys, and CRMs, extracts insights with the customer's quote, and groups them into Themes. Unified contacts match people by email and by Slack, Intercom, Zendesk, Salesforce, HubSpot, and Gong IDs, so one customer posting in five places counts once.
  • Segments and breadth: rules tag feedback by email domain, plan, support-tool tags, Slack channel, and more (Deck segments docs). The weekly Impact score, from 0 to 10, counts source diversity and customer breadth, and scoping an opportunity to a segment re-scores it for those customers (Opportunity Backlog).
  • ARR and churn risk: with a CRM connected, on Business and above, you can filter Themes and insights by fields such as ARR or churn risk (Deck Salesforce docs).
  • Limitations: revenue and churn risk are not inputs to the Impact score, and you can't set custom weights, so the capped-ARR math lives in a sheet or with an AI agent on Deck's MCP server (Deck workflow docs). The default backlog hides opportunities with fewer than three pieces of evidence. There is no public voting board, so keep yours and import its posts by CSV.

To try it, the Free plan includes 200 AI credits a month and one integration; Business, which adds CRM integration, is $249 a month or $199 billed annually (Deck pricing).

Frequently Asked Questions

Usually not. Keep it for intake, for the context in its comments, and as a list of people to tell when something ships. Just stop using its vote count as the ranking.

Cap each account's ARR at roughly your 90th-percentile contract before you add it up. A large customer then counts several times as much as a median one but cannot outweigh dozens of smaller accounts with the same problem.

Count it only when an account is at risk and names the request, or the problem behind it, in a renewal call, cancellation survey, escalation, or CSM health note. A red health score alone does not tie the risk to any one request.

Give each free account an expected value instead of zero or full weight: your median paid ARR times your free-to-paid conversion rate. With a $12,000 median and 4% conversion, each free account counts as $480.

Re-run the score monthly with fresh ARR, confirmed churn flags, and the last 90 to 180 days of requests. Reset the weights and strategic fit once a quarter at planning, and leave them alone mid-quarter unless the strategy changes.

Sources

Sources checked November 2, 2026.