The discovery backlog is a rolling recruit pipeline — sourced, invited, scheduled, done — that keeps at least two customer interviews booked at all times, so a single canceled call or a brutal sprint never leaves your weekly discovery slot empty. It runs on a self-serve scheduling link and a standing in-product ask, not on you manually chasing names every Friday afternoon.
Quick Answer: Keep a backlog of candidates moving through four stages (sourced → invited → scheduled → done) and never let "scheduled" drop below two. A self-serve booking link plus an automatic in-app prompt keeps the pipeline fed without you having to recruit from scratch each week.
Why the weekly habit breaks on the recruiting step, not the interviewing step
The failure point in continuous discovery is almost never the conversation itself — it's the 20 minutes before it that nobody scheduled. Teresa Torres's weekly discovery habit of at least one customer touchpoint per week is well understood, but the operational gap is that most teams treat recruiting as a task they do when they need an interview, not a pipeline they maintain independent of need.
That's backwards. If you start looking for a participant on Monday for a Thursday interview, you're one canceled meeting, one unresponsive contact, or one busy sprint away from a blank week. And a blank week is where the weekly discovery habit quietly dies — not through a decision to stop, but through a string of weeks where nobody had anyone booked.
The fix is to decouple sourcing from scheduling. Recruiting becomes a standing background process — always running, always a few candidates ahead of where you actually need them — so the interview slot is filled by the time you get there, not scrambled for.
The math behind "always-two-booked"
One booked interview is a single point of failure. Customers reschedule, no-show, or turn out to be a poor fit mid-call. Two booked interviews at all times gives you redundancy without over-recruiting: if one falls through, you still have a session, and you have slack to debrief and update your assumptions before the next one arrives.
| Booked interviews | What happens when one cancels | Recruiting overhead |
|---|---|---|
| Zero (recruit on demand) | Empty week; habit breaks | Highest — full search under time pressure |
| One | Empty week if it cancels/no-shows | Moderate — constant low-grade urgency |
| Two (target state) | Week still holds | Low — steady background sourcing |
| Four+ | Rarely helps discovery quality; risks stale insights before you talk to people | Higher than needed — recruiting effort better spent researching new segments |
Two is a target, not a hard ceiling — some weeks you'll have three or four in flight because candidates convert unevenly. The point is a floor, not a cap: never let scheduled interviews fall below two.
Build the pipeline: four stages, one shared view
A discovery backlog needs exactly four stages to function — sourced, invited, scheduled, done — each with a clear entry and exit condition. Skipping a stage (e.g., inviting people you haven't actually sourced with intent) is usually where pipelines silently stall, because nobody notices the backlog thinning until the week it's already empty.
- Sourced — a name and contact method exists, tagged with why they're relevant (segment, behavior, recency of a support ticket, a
jobs-to-be-donesignal). Nothing has been sent yet. - Invited — an outreach message has gone out with a scheduling link. This stage has a clock: if no response in 3-4 business days, follow up once, then recycle the contact back to "sourced" for a different ask later.
- Scheduled — a calendar slot is confirmed. This is the stage you're protecting at a floor of two.
- Done — the interview happened; notes are logged and linked to the relevant assumption or opportunity in your opportunity solution tree.
Where sourced candidates actually come from
A backlog can't run on goodwill alone; it needs a repeatable sourcing mechanism, or "sourced" stays empty and the rest of the pipeline has nothing to pull from. Rotate across a small number of channels so you're not draining a single well:
- In-product prompts triggered by specific behaviors (recent cancellation, heavy usage of a feature you're investigating, a completed onboarding flow).
- Support and sales handoffs — anyone who mentioned friction in a ticket or a deal call is a warmer lead than a cold list.
- A standing signup form linked from your product, positioned as "help shape what we build," not a generic survey.
- Existing customer advisory or beta lists, rotated so the same five people aren't interviewed every month.
Weight sourcing toward the segment tied to your current opportunity tree, but keep 10-20% general — discovery that only talks to people who already match your hypothesis tends to confirm what you already believed instead of surfacing what you missed.
The self-serve scheduling link removes the coordination tax
A shared scheduling link (Calendly, SavvyCal, or equivalent) turns "find a mutually available half hour" from a multi-email negotiation into a single click, which is the single biggest lever for moving candidates from invited to scheduled quickly. Without one, the invited stage becomes the pipeline's actual bottleneck — not because people don't want to talk, but because back-and-forth scheduling has too much friction for a low-commitment ask.
Set the link up with intent, not as a generic "book time with me" page:
- Cap slots to your real cadence — if you only want two interviews live at once, don't expose ten open slots a week; expose three or four so the invite pool doesn't overcommit you.
- Buffer 10 minutes before and after every slot for notes and context-switching — a back-to-back interview schedule produces worse notes than a spaced one.
- Pre-qualify with two or three fields on the booking form (role, how long they've used the product, what they're trying to do) so you walk in with context and can steer toward open, non-leading questions instead of guessing who's on the call.
- Auto-send a confirmation with a light prompt ("we'll mostly talk about how you currently handle X") so participants arrive primed without being coached toward an answer.
The automatic in-product ask
The highest-conversion invite usually isn't an email at all — it's a well-timed in-app prompt shown to someone actively using the relevant feature, because intent and context are both fresh in that moment. An automatic ask embedded at a specific trigger point (after a workflow completes, after a support interaction resolves) converts noticeably better than a cold outreach email to the same list, because the person is already in the mental context you want to discuss.
Keep the ask short and honest about time commitment — "15 minutes, this week, about how you just did that" — and link straight to the scheduling page with the context pre-filled. Treat this as one channel among several, not a replacement for sourcing intentionally from segments the in-app trigger can't reach (churned users, prospects, non-users of the feature in question).
Protecting the habit: what the backlog actually buys you
A maintained backlog converts discovery from a willpower exercise into a systems problem, which is the entire point — habits that depend on remembering to do something under time pressure fail exactly when time pressure is highest. The backlog exists precisely for the sprint where nobody has bandwidth to recruit; that's when the buffer you built in the calm weeks pays off.
Run a simple weekly check, ideally as a recurring calendar block, not an ad hoc glance:
| Stage | Healthy floor | Action if below floor |
|---|---|---|
| Sourced | 5+ names in queue | Pull from a sourcing channel above; don't wait until invited runs dry |
| Invited | 3+ awaiting response | Send this week's outreach batch now, not Thursday |
| Scheduled | 2+ confirmed | Escalate: personal follow-up to warmest invited contacts |
| Done (rolling 4 weeks) | 1+ per week, no gaps | Retrospective: which stage is actually leaking |
If "scheduled" ever drops below two, that's the signal to act — not the empty calendar on the day you needed the interview. Treating the floor as a leading indicator, not a lagging one, is what actually protects the weekly discovery habit through busy periods instead of merely diagnosing why it broke after the fact.
Tying interviews back to what you're actually deciding
A backlog that produces interviews disconnected from your current opportunity space is busywork with good attendance. Each "done" interview should map to a node on your opportunity solution tree or a stage in your customer journey — otherwise you accumulate transcripts without accumulating decisions. Jim Kalbach's jobs-to-be-done framing and Bob Moesta's Forces of Progress model are both useful lenses for making sure each conversation is anchored to a real job the customer is trying to get done, not just a friendly chat. For teams new to structuring this end-to-end, the complete guide to product discovery covers how weekly interviews, opportunity mapping, and prioritization fit together as one system rather than three disconnected habits.
Keeping the pipeline alive under real sprint pressure
Recruiting is the first thing that gets dropped mid-sprint because it doesn't feel urgent the week it's skipped — it only feels urgent two weeks later when the calendar is empty. The backlog's stages need a recurring nudge that survives a team's attention being pulled elsewhere, since a spreadsheet nobody checks is functionally the same as no pipeline at all.
This is the honest, narrow use case for a recurring reminder layered on top of the pipeline you already run: Prodinja's Reminders are designed to sit on a recurring beat — a weekly check of the sourced/invited/scheduled counts, a nudge to send this week's outreach batch, a prompt to log the last interview's notes before details fade. It's a scheduling assist for a process you already own, not a substitute for actually doing the recruiting; the pipeline mechanics above are what make the reminder worth having in the first place.
Key Takeaways
- Recruiting is a pipeline, not a task — track sourced, invited, scheduled, and done as distinct stages with clear entry and exit conditions, not one undifferentiated to-do.
- Always-two-booked is the floor to protect, not a target to hit occasionally — one booked interview is a single point of failure against a cancellation or no-show.
- A self-serve scheduling link removes the coordination tax that stalls candidates in the invited stage; cap slots to your real cadence and buffer time around each one.
- An in-product ask at the right trigger point often converts better than cold outreach because the participant is already in the relevant mental context.
- Check the backlog weekly against floors, not just when a week comes up empty — scheduled dropping below two is the leading indicator, not the empty calendar itself.
- Every interview should map to a live opportunity or journey stage — a full pipeline that isn't tied to current decisions produces transcripts, not insight.
- A recurring reminder protects the habit through busy sprints by keeping the recruiting beat alive even when discovery isn't the loudest priority that week.
Frequently Asked Questions
How many people should be in a discovery backlog at once?
A healthy backlog typically holds 5+ sourced names, 3+ invited and awaiting response, and at least 2 confirmed scheduled interviews at any given time. These are floors, not caps — some weeks will run higher as candidates convert unevenly through the stages.
How do I recruit interview participants without an existing customer list?
Combine in-product prompts triggered by relevant behavior, support and sales handoffs from teams already talking to customers, and a standing signup form positioned as shaping the roadmap. Rotate across channels so no single source dries up and stalls the entire pipeline.
What's the difference between a discovery backlog and just scheduling interviews as needed?
Scheduling as needed makes recruiting reactive — you start searching the week you need someone, which means a single delay leaves the slot empty. A backlog keeps sourcing running continuously in the background, several candidates ahead of when you'll actually need them.
How often should I follow up with an invited candidate before giving up?
One follow-up after 3-4 business days of silence is a reasonable default before recycling the contact back to "sourced" for a different ask later. Chasing much longer than that usually costs more attention than it returns, especially when fresh candidates are entering the pipeline anyway.
Does a scheduling link replace personal outreach for high-value participants?
No — a self-serve link handles volume and removes coordination friction, but your most strategic candidates (a churned enterprise account, a highly specific segment) still warrant a personal message. Use the link as the default path and reserve manual outreach for candidates where the extra effort is clearly worth it.