Continuous decisioning replaces the weekly promo calendar with a loop that ships hundreds of small, causal experiments per week against a preserved holdout. A 90-day migration.
Frequently asked
Related resources
Solutions
See it in the product
The same loops this note describes run 24/7 against your customer base. Watch the workspace decide, experiment and execute 1:1.
Vocabulary
Definitions an answer engine can quote, each one a page of its own.
Compare
Neutral reads on the categories buyers evaluate against a decision layer.
All comparisonsAlmost every large B2C growth team we meet still runs on a calendar. A weekly planning meeting, a monthly promo grid, a quarterly campaign roadmap. The artifacts are different across streaming, telecom, fintech and retail, but the operating model is identical: batch the decisions, ship them on a cadence, read the results after the fact. That cadence is the single largest source of trapped ARPU in the enterprise, and it is not obvious why until you look at what the cadence is actually costing.
The alternative is not “always-on campaigns.” That is the same calendar with fewer gaps. The alternative is continuous decisioning: a system that treats every customer moment as a decision, generates hundreds of candidate actions in parallel, and reads each one causally. This piece is a practical guide to that shift, from why the calendar is holding you back to how to move off it in ninety days without breaking the business.
The weekly promo calendar is a legacy of paid media buying, where inventory had to be reserved in advance and creative had to be produced against a fixed slot. Owned channels inherited the cadence even though the underlying constraint had vanished. Email, push, in-app, on-site and outbound are all instantly addressable at any moment, to any subset of the base, with any variant. The calendar is not a property of the channel. It is a property of the team.
That would be fine if the cost were only aesthetic. It is not. Batching decisions on a weekly cadence forces three specific compromises, and each of them destroys measurable ARPU.
A promo shipped to a segment of two million people is really forty small experiments glued together. Some subsegments respond with meaningful lift, others are indifferent, a few are actively cannibalized. When the whole batch ships as one decision, the readout is one number, and that number is a weighted average. The interesting variance is gone before anyone sees it. You learn that the campaign “worked,” which is not a learning at all.
Under a calendar, the unit of work is a campaign, and the unit of a campaign is a creative brief. Everything the growth team argues about is downstream of that brief: subject lines, images, offer strength. The segment, the moment and the counterfactual, which is where the actual ARPU delta lives, become afterthoughts of the creative. The best-performing creative on the wrong segment is worth less than a mediocre creative on the right one.
The most reliable pattern we see: when a calendar slips, the first thing that gets cut is the holdout. Reach targets are contractual with the commercial team, holdouts are not. The team ships the send to 100% of the segment, tells the CFO the revenue number, and quietly loses the ability to prove incrementality. Two quarters of this and the whole retention program is running on faith.
Continuous decisioning replaces the calendar with a loop. The loop has six stations and it never stops running.
Nothing about that loop is exotic. Every good growth team has run each of those steps individually. What changes under continuous decisioning is the volume: instead of four campaigns per month, the loop is running four hundred experiments per month, each one small, each one scoped, each one readable in isolation.
Consider two operating models, both aimed at the same base of ten million active customers.
Model A, the calendar. Four campaigns per month. Average lift, when measured with a proper holdout, is five percent on the addressed segment. Each campaign reaches roughly one and a half million people. Twelve months of this produces an incremental ARPU contribution measured in low single digits, most of which is not causally attributed.
Model B, continuous decisioning. Four hundred micro-decisions per month. Average lift is two percent per decision, on a tightly scoped segment of thirty to fifty thousand customers, with a holdout preserved. Individually each experiment is small. Compounded across the base, with the losses discarded and the winners scaled, twelve months of this reliably moves total ARPU by a mid-single-digit percentage, and every dollar of it is auditable.
The reason Model B wins is not that the lifts are larger. They are smaller. It wins because the base of decisions is two orders of magnitude larger, the holdouts are preserved so the winners are trusted, and the losers are cut fast enough that the portfolio drifts toward its best performers on its own. Decision volume, not creative volume, is the multiplier.
Continuous decisioning does not require a bigger growth team. It requires a differently pointed one. The activities that used to consume the week, campaign briefing, calendar negotiation, send QA, post-mortem decks, either disappear or collapse into a review tab. In their place, four activities expand.
The teams that do this well look less like a CRM function and more like a small research group. That is not an accident. We wrote more about the shape of that team in What a top data science team looks like in an agentic era, which will follow this piece.
The mistake most teams make is to try to retire the calendar everywhere at once. Do not. Pick one segment, one channel and one metric, and run the new operating model against them in parallel with the existing calendar. Ninety days is enough to prove the point.
Freeze the calendar for one segment. Instrument a real holdout, ten to twenty percent, and defend it against every request to backfill. Ship twenty candidate actions in parallel against small subsegments, each with a written hypothesis. At day 30 you should be able to point to at least three causally validated winners and at least five cleanly retired losers. If you cannot, the instrumentation is wrong, not the strategy.
Take the validated winners and expand them to adjacent segments, still with holdouts. Add a second channel to the loop. Start writing hypotheses at three times the previous volume. This is where the org shift becomes visible: the weekly meeting stops being a calendar review and becomes an experiment review.
By day 90 the segment under the new model should be producing a measurable ARPU delta relative to the segments still on the calendar. The delta is the argument. Present it to the CFO with the holdout methodology attached and use it to fund the migration of the next segment. From here the program self-funds.
Continuous decisioning surfaces failure modes that a calendar hides. Four in particular are worth naming up front, because every serious migration hits them.
A year into a serious migration, the growth team is unrecognizable. There is no promo calendar on the wall. There is a live opportunity feed, an experiment board, and an incrementality dashboard. The weekly meeting is thirty minutes long and it argues about hypotheses, not sends. The CFO has stopped asking for ROI justifications on individual campaigns because the incrementality report is a standing artifact.
More importantly, the ARPU curve has changed shape. It stops looking like a step function that jumps around each launch and starts looking like a line that compounds. That is the entire point. The calendar produces events. Continuous decisioning produces a rate.
For the underlying data on how the leaders are already operating this way, see The 2026 ARPU Report for B2C Enterprises.
