Booking curve analysis — how a date fills
In lesson 17 (Booking pace and the pace report) we learned to read pace as a curve — the time trajectory of how occupancy builds. Now we deepen that curve into a date-level fingerprint: the booking curve shows how occupancy has historically built up for a given day type. This fingerprint becomes the basis of the forecast and the yardstick for anomaly detection.
How a Saturday fills is not random — it reflects the hotel’s own deep demand pattern. If nearly every Saturday of the past three years built along a similar path, the next Saturday will most likely follow the same path — that is the statistical basis of curve-based forecasting. And it works in reverse too: when a date departs from its usual path, that almost always means something.
The goal of this lesson is for you to understand: what the booking curve is, what shapes it takes, how to spot a curve deviation, and how the curve becomes a pace-adjusted forecast.
What the booking curve is
The booking curve (in the industry literature it also goes by fill curve) is the average historical build-up path of a given day type: at what percentage OTB stands 90, 60, 30, 14, 7, 3 days before arrival — on average. Hotel Peaqplus City’s (80 rooms) Saturday curve, averaged from the past 3 years of Saturdays (with event days filtered out — more on that later):
| Days to check-in | 3-year average OTB | Average daily pickup to the next point |
|---|---|---|
| 90 days | 8% | +0.2 rooms/day |
| 60 days | 15% | +0.3 rooms/day |
| 45 days | 21% | +0.6 rooms/day |
| 30 days | 32% | +0.9 rooms/day |
| 21 days | 42% | +1.4 rooms/day |
| 14 days | 54% | +1.6 rooms/day |
| 7 days | 68% | +2.4 rooms/day |
| 3 days | 80% | +2.8 rooms/day |
| 1 day | 87% | +4.0 rooms/day (last-minute + walk-in) |
| 0 days (final) | 92% | — |
The pickup column can be verified against the 80 rooms: 1 pp = 0.8 rooms, so e.g. the +14 pp of the 14→7 day stretch is 11.2 rooms over 7 days = +1.6 rooms/day. This is the booking velocity — the speed at which bookings arrive on that stretch.
The shape is the classic S-curve: a slow start (90-45 days), an accelerating middle (30-7 days), then the last-minute + walk-in peak of the closing days that carries the curve to its 92% final (≈ 74 rooms). If the numbers look familiar: in lesson 17 we used exactly this historical pace path for the pace vs. budget comparison (32% at 30 days, 54% at 14, 68% at 7) — the booking curve is that concept extended across the full booking window.
The booking curve serves two purposes:
- Forecast basis — we measure the current pace position against the same point of the curve and project the likely final from it.
- Anomaly signal — when the current pace deviates significantly from the curve, it almost always points at something concrete: an event, a campaign effect, softening demand.
The three main curve shapes
From lesson 17 you know the five pace shapes (S-curve, frontload, backload, flat, double-peak). At curve level, the three main base types are worth knowing more deeply — each is a mirror of the segment mix.
1. Classic S-curve (mixed segment mix)
The Saturday curve above. Mixed mix: the bulk of leisure books 45-14 days ahead, business and corporate in the last two weeks, walk-ins on the day. This is Hotel Peaqplus City’s healthy weekend base pattern — and the most common curve shape in most city hotels.
2. Frontload curve (event-driven)
The curve runs high early — 60 days before arrival OTB is already at 40-50%. Guests know about the event in advance and book early. Hotel Peaqplus City’s arena-concert Saturday last year (a Coldplay show) looked like this against the standard Saturday:
| Days to check-in | Standard Saturday | Event Saturday |
|---|---|---|
| 90 days | 8% | 22% |
| 60 days | 15% | 42% |
| 30 days | 32% | 68% |
| 14 days | 54% | 82% |
| 7 days | 68% | 92% |
| 0 days (final) | 92% | 98% |
Sixty days out, the event Saturday runs +27 pp ahead of standard — 43% of its 98% final is already on the books, while the standard Saturday sits at only 16% of its 92% final. The other side of the ledger: in the last two weeks the event date adds only +16 pp, against the standard’s +38 — most of the demand booked months earlier. Letting a day like this fill at standard rates throughout means admitting the early, price-insensitive demand wave too cheaply — by the dynamic-pricing logic of lessons 35 and 36, the right answer here is an early rate increase.
3. Backload curve (corporate-driven)
The opposite: a slow start (30 days out often only ~20% OTB), then a fast close in the final week. Hotel Peaqplus City’s Tuesday curve:
| Days to check-in | Tuesday curve |
|---|---|
| 30 days | 20% |
| 14 days | 30% |
| 7 days | 48% |
| 3 days | 66% |
| 0 days (final) | 74% |
Tuesday still stands at just 20% thirty days out, yet the final is 74% — nearly three-quarters of the occupancy arrives in the last 30 days, of which +26 pp in the final week, at a 3-4 rooms/day pickup in the 7-3 day window. This is the signature of short-lead corporate and transient business — in lesson 9’s weekly pattern we saw that Tuesday-Wednesday at Hotel Peaqplus City is business-driven. On a backload day, 20% at 30 days is no cause for panic: that is the normal path. The cause for panic is the last-week pickup wave failing to arrive.
Curve deviation as an anomaly signal
The curve’s real power is the daily comparison: the current pace position against the same point of the curve.
Example 1 — the slow Saturday
Hotel Peaqplus City’s Saturday, Dec 9, at the morning pace check:
| Days to check-in | Standard curve | This year’s Dec 9 | Gap |
|---|---|---|---|
| 14 days | 54% | 44% | −10 pp |
| 10 days | 62% | 48% | −14 pp |
| 7 days | 68% | 52% | −16 pp |
Two things are visible. First, the date is behind the curve. Second — and this matters more — the gap is growing: −10 → −14 → −16 pp. The curve would expect +14 pp of pickup across the 14→7 day stretch; +8 actually came. The date is not catching up, it is slipping further.
The final projectable from the curve (we derive the calculation in the forecast section): still ~75% at 14 days, only ~70% by 7 days — instead of the usual 92%. Action is needed: a downward BAR revision, a targeted promotion (with lesson 31’s timing logic), a channel-visibility check. And one control question from lesson 18 before you cut rates: is there a structural cause — an event last year that does not repeat this year?
Example 2 — the too-fast Saturday
Hotel Peaqplus City’s Saturday, Nov 11:
| Days to check-in | Standard curve | This year’s Nov 11 | Gap |
|---|---|---|---|
| 21 days | 42% | 56% | +14 pp |
| 14 days | 54% | 68% | +14 pp |
The date runs a steady +14 pp ahead of the curve. The proportional projection runs above 100% (92 × 68/54 ≈ 116%, which in reality caps at 100%) — the date is heading toward selling out, well above its usual level. What might it signal?
- Event demand the hotel does not know about — a compset-shopping check (lesson 32): have competitor rates jumped for the same night?
- A campaign that works — the marketing action is delivering the volume.
- A larger connected block of bookings (e.g. a wedding) that arrived as transient reservations.
The action here points the other way: a rate increase and a rethink of restrictions while sellable capacity remains. A slow anomaly calls for a demand action, a fast anomaly for a price action — this is the most important decision rule of curve analysis.
The segment-level booking curve
A mature RM organisation reads the curve by segment too. The standard Saturday curve broken down — the four segment columns sum to the total curve in every row:
| Days to check-in | Transient leisure | Transient business | Corporate | Group (definite) | Total |
|---|---|---|---|---|---|
| 90 days | 3% | 0% | 0% | 5% | 8% |
| 60 days | 8% | 0% | 0% | 7% | 15% |
| 30 days | 21% | 1% | 0% | 10% | 32% |
| 14 days | 36% | 3% | 3% | 12% | 54% |
| 7 days | 44% | 7% | 5% | 12% | 68% |
| 3 days | 52% | 10% | 6% | 12% | 80% |
| 0 days (final) | 58% | 13% | 9% | 12% | 92% |
The breakdown is telling:
- Transient leisure delivers nearly two-thirds of the final (58 pp of the 92), and the bulk of its build-up happens in the 60-7 day band.
- Transient business (13 pp) arrives practically in the last two weeks.
- Corporate (9 pp) trickles in on even shorter lead times, in the final days.
- Most of the group definite volume is on the books 90-60 days ahead, and no longer moves inside 14 days.
The segment-level curve gives a more precise diagnosis than the total. If transient leisure stands at only 12 pp thirty days out instead of the usual 21, that is a leisure-specific problem — campaign, visibility, pricing on the leisure channels — not a general demand slump. And the total curve can mask it: it happens that the total sits exactly on the curve while the composition has shifted — and a mix shift carries an ADR consequence too (lesson 21’s mix analysis).
The booking curve as a forecast basis
In lesson 20 (Simple forecast methods), Layer 2 of the 3-layer forecast was the pace extrapolation. The booking curve is its main input. The most common form is the proportional (pace-adjusted) projection:
Projected final = curve final × (current OTB / curve level at the same days-out point)
On the Dec 9 example, 14 days before arrival:
- Current OTB: 44%
- Curve level at 14 days: 54%
- Curve final: 92%
- Pace-adjusted projection: 92% × (44/54) ≈ 75%
The plain curve average would say 92% — the pace-adjusted projection says 75%. The 17 pp difference is exactly the information the current pace carries: the date is running below its usual path, and the final estimate has to reflect that.
Try the calculator below — it computes the other method you know from lesson 20, the incremental (same-point based) projection: base final + pp gap. With the Dec 9 numbers: 92 + (44 − 54) = 82%. The proportional method carries the shortfall through proportionally (75%), the incremental keeps it in pp (82%) — a mature model combines the two with weights, here landing around 78%. The exact number depends on the method; the message does not: the 92% average is not going to happen.
Mature forecast models — including Peaqplus Smart Forecast — never work off the plain curve average: they are always pace-adjusted, and they recompute the projection every day, for every date.
The difficulties of calibrating the curve
The booking curve is not an eternal truth. Four typical maintenance problems:
1. Data ageing
The 3-year average can go stale when the market changes underneath it — a new compset hotel opens, the demand structure rearranges. The established fix is a weighted average: last year at 50%, two years ago at 30%, three years ago at 20%.
2. Separating event days
Event days must not be mixed into the average: a festival week or an arena-concert Saturday runs on a different path and would distort the standard curve upward. Mature practice keeps a separate event curve and standard curve — which is why the very first table said: event days filtered out.
3. Segment shift
If the hotel’s mix changes structurally (e.g. a new MICE business line builds up), the old curve is no longer the hotel’s fingerprint. The curve then has to be re-based — on a shorter historical window, with more weight on fresh data.
4. Calendar shift
Familiar from lesson 18 (Same point analysis): last year’s November 25 fell on the Black Friday weekend, this year’s does not — the same date, a different calendar role. The curve deviation in such cases is a structural cause, not an RM error; it calls for comparable-date matching, not an action.
Peaqplus Pickup + Smart Forecast in curve analysis
Done manually, curve building is painful work: reservation exports, date-by-date retrospective OTB reconstruction, constant maintenance. In Peaqplus this layer is built in.
The Pickup module shows three columns for every future date on the pickup board: the current OTB, the curve level (the hotel’s own historical path at the same days-out point) and the colour-coded gap — green for dates running ahead of the curve (rate-increase candidates), yellow for a small shortfall (watch), red for a significant slip (action needed). Daniel’s morning routine is thus not curve building but curve reading: from the 90-day board, the anomalies jump out in 2-3 minutes.
Smart Forecast uses the curve position in the three-layer stack you know from lesson 20: Layer 1 is the comparable-date-matched historical base, Layer 2 is the extrapolation of the pace position measured against the curve, Layer 3 is the manual corrections (event, calendar shift, group). The layers combine with weights, and the model attaches a confidence range to the estimate. In lesson 38 (Smart Forecast — hybrid forecasting) we go through its inner workings.
Key takeaways
- The booking curve is the average historical build-up path of a day type — a date-level fingerprint, the shared basis of forecasting and anomaly detection.
- The three main curve shapes mirror the segment mix: S-curve (mixed mix), frontload (event — high OTB early), backload (corporate — fast late close).
- A curve deviation is instant diagnosis: a slow anomaly (growing gap) calls for a demand action, a fast anomaly for a rate increase.
- The segment-level curve is more precise than the total: the total can mask a mix shift — and its ADR consequence.
- The pace-adjusted projection (curve final × current OTB / curve level) is a far more realistic estimate than the curve average — 75% instead of 92% in the Dec 9 example.
- The Peaqplus Pickup module shows the curve position daily, colour-coded; Smart Forecast builds it into the forecast automatically.
Click an answer — you see immediately whether it is right.
Answer all of them and the lesson counts as complete — and toward your progress.
Pace gap = this year OTB − last year same-point. Projected final = last year final + pace gap.
See the full definitions in the glossary.
A hotel's booking curve for Mar 8 (Friday): 30% thirty days before arrival, 50% at 14 days, 70% at 7 days, final 88%. This year's Mar 8 stands at 65% today, 14 days out. What is your proportional (pace-adjusted) projection, how do you interpret the result, and which three possible explanations would you investigate? And: Hotel Peaqplus City's Sunday curve sits at 35% fourteen days out with a 58% final. What DCAL day-class assignment and what restriction / promotion strategy would you build for this Sunday shape? Walk it through.
- Measuring curve deviation is an industry standard: STR's pace reports compute the same logic as a "pace index" at market level. An independent RM reconstructs booking curves from reservation exports in Excel — hours of work per date that a mature RMS recomputes automatically every day.
- In the classic revenue management literature the booking curve also goes by "fill curve" — the concept arrived in hotels from airline yield management, where the flight-fill curve is the basis of demand forecasting in exactly the same way.