Drilling Time Estimation - From ROP Calculations to Predictive Modeling
Drilling time estimation is where well planning meets financial accountability. Every AFE (Authorization for Expenditure) is anchored to a drilling time estimate, and every day of overrun costs $50,000 to $500,000 depending on the rig type. Yet drilling time estimation remains one of the least rigorously practiced disciplines in well planning - most estimates are based on "what we drilled last time" without systematic analysis of what drove that performance and how the new well differs. This guide gives you the complete framework: the physics-based ROP models, the statistical methods for using historical data, and the practical tools that separate accurate estimates from expensive guesses.
1. Rate of Penetration - The Foundation of Drilling Time Estimation
1.1 The Basic Drilling Time Calculation
At its simplest, drilling time for any interval is:
Rotating time (hours) = Interval length (ft) / ROP (ft/hr)
Total section time = Rotating time + Connection time + Survey time + Wiper trips + NPT allowance
Connection time = (Interval / 30ft stand length) x (3-5 min/connection) / 60
Survey time (MWD) = (Interval / 90ft survey interval) x (5-10 min/survey) / 60
Worked example - 8.5" hole section, 3,000 ft interval:
- Average ROP = 65 ft/hr
- Rotating time = 3,000 / 65 = 46.2 hours
- Connections = (3,000/93ft) x 4 min / 60 = 32 connections x 4 min = 2.1 hours
- MWD surveys = (3,000/90ft) x 8 min / 60 = 33 surveys x 8 min = 4.4 hours
- NPT allowance (8%) = (46.2 + 2.1 + 4.4) x 0.08 = 4.2 hours
- Total section time = 56.9 hours = 2.4 days
1.2 Factors That Control ROP
ROP is not a single number - it varies with every parameter change during drilling. Understanding what drives ROP allows you to predict it for new wells rather than simply copying offset well performance:
| Factor | Effect on ROP | Quantified Relationship |
|---|---|---|
| Weight on bit (WOB) | Increases ROP linearly until bit flounder point | ROP doubles as WOB doubles (below flounder point) |
| RPM | Increases ROP linearly for roller cone, less so for PDC | 10% RPM increase = ~7-10% ROP increase for RC bits |
| Formation compressive strength (UCS) | Higher UCS dramatically reduces ROP | ROP inversely proportional to UCS^0.5 to UCS^1.0 |
| Hydraulic horsepower at bit (BHHP) | Improves cuttings removal from under bit | Optimal BHHP > 3.5 HP/in2 bit area |
| Differential pressure (overbalance) | Higher overbalance reduces ROP (chip hold-down effect) | Every 500 psi overbalance reduces ROP by 5-15% |
| Bit wear (dull grade) | Progressive ROP decline as bit dulls | ROP at dull grade 7/8 typically 40-60% of fresh bit ROP |
1.3 The Bourgoyne-Young ROP Model
The most widely used physics-based ROP model in the industry combines all major driving parameters:
ROP = f1 x f2 x f3 x f4 x f5 x f6 x f7 x f8
Where each fi is a function factor:
f1 = formation strength effect (exponential with depth)
f2 = normal compaction trend deviation
f3 = pressure differential effect
f4 = bit weight effect: f4 = ((WOB/db - (WOB/db)_t) / (4 - (WOB/db)_t))^a5
f5 = RPM effect: f5 = (N/60)^a6
f6 = bit tooth wear effect
f7 = hydraulics (bit jet impact) effect
f8 = bit type factor
Where db = bit diameter (inches), N = RPM, a5 and a6 are regression constants from offset wells
Practical application: The B-Y model requires calibrating the regression constants (a1 through a8) from offset well data. Once calibrated to a field dataset of 3-5 wells, it typically predicts ROP within 15-25% accuracy for new wells in the same formation. Without calibration, treat it as a qualitative tool only.
2. Using Historical Data - The Right Way
2.1 Building an Offset Well Database
Raw offset well data is rarely directly applicable to a new well estimate. Before using any historical ROP data, it must be normalized for the key variables that differed between wells:
| Variable to Normalize | Why It Matters | Normalization Method |
|---|---|---|
| WOB and RPM | Different drilling parameters give different ROP even in same formation | Scale ROP to a standard WOB/RPM using B-Y factors |
| Bit type and size | PDC vs roller cone drill at very different ROP in same formation | Separate datasets by bit type, apply bit efficiency factor |
| Mud weight | Higher overbalance suppresses ROP via chip hold-down | Apply differential pressure correction factor |
| Pump rate (hydraulics) | Poor bit cleaning reduces ROP independently of formation | Normalize to standard BHHP/in2 |
| Bit hours run | Worn bit gives lower ROP in same formation | Use only fresh bit ROP data or apply wear correction |
2.2 Statistical Analysis of Offset Data
Once normalized, offset well ROP data should be analyzed statistically rather than simply averaged:
P50 estimate = median ROP from normalized offset dataset
P10 estimate (optimistic) = 90th percentile ROP
P90 estimate (conservative) = 10th percentile ROP
Section drilling time P50 = Interval / ROP_P50
AFE time = P50 drilling time + (P90 - P50) x risk factor
For frontier wells with limited data: use P90 as base case for AFE
Example with 5 offset wells in the same formation interval (normalized ROP in ft/hr):
| Well | Raw ROP (ft/hr) | Normalized ROP (ft/hr) | Notes |
|---|---|---|---|
| Offset-1 | 85 | 72 | High WOB run - scaled down |
| Offset-2 | 61 | 68 | Low mud weight - scaled up slightly |
| Offset-3 | 52 | 65 | Worn bit - corrected for wear |
| Offset-4 | 78 | 75 | Good conditions - minimal correction |
| Offset-5 | 44 | 48 | Hard stringer encountered - representative |
| Statistics | Mean: 64 | P50: 68, P10: 75, P90: 48 | Use P50=68 for base, P90=48 for conservative AFE |
2.3 The Time vs Depth Plot - Your Primary Planning Tool
Plot cumulative well time on the X-axis against depth on the Y-axis for all offset wells on the same chart. This gives you:
- Visual identification of problem zones - any depth range where offset wells suddenly became nearly horizontal (slow drilling or NPT) is a risk zone
- P10/P50/P90 time envelopes by connecting the fastest, median, and slowest depth-time curves
- NPT identification - flat sections on the time-depth plot are NPT events; quantify them separately from drilling time
- Casing point timing - identify at which depth each offset well set casing and how long the casing job took
3. Building the Complete Well Time Estimate
3.1 Activity-Based Time Breakdown
A rigorous drilling time estimate breaks the well into activities, not just footage. Each activity has its own time driver:
| Activity Category | Time Driver | Typical % of Well Time |
|---|---|---|
| Drilling (rotating) | ROP x interval length | 30-45% |
| Tripping (in and out) | Depth / trip speed (typically 1,000-1,500 ft/hr) | 15-25% |
| Casing and cementing | Number of strings x time per job (typically 12-24 hrs) | 10-15% |
| Logging (wireline/MWD) | Interval / logging speed (typically 1,800-3,600 ft/hr) | 5-8% |
| BHA changes and inspections | Number of BHA runs x rig-up/rig-down time (2-4 hrs) | 3-5% |
| Well control and kicks | Statistical - based on offset well frequency | 1-3% |
| NPT (stuck pipe, lost circ, etc.) | Historical NPT% for operator/area/formation | 8-20% |
3.2 NPT - The Most Underestimated Component
NPT (Non-Productive Time) is where most AFE overruns originate. Industry average NPT is 15-20% of total well time, but most drilling time estimates use 5-8%. The discrepancy arises because planners use best-case NPT from their best offset wells, not the statistical average.
NPT should be estimated by category, not as a single percentage:
| NPT Category | Industry Average Frequency | Average Duration | Mitigation |
|---|---|---|---|
| Stuck pipe | 1 per 3-5 wells | 12-48 hours | Mud properties, BHA design, reaming program |
| Lost circulation | 1 per 2-4 wells | 6-24 hours | ECD management, LCM pre-treatment |
| Equipment failure (top drive, pumps) | 1-2 per well | 4-12 hours | Maintenance schedule, spare parts inventory |
| Wellbore instability (tight hole, reaming) | Depends on shale content | 2-8 hours per occurrence | Inhibitive mud, optimized mud weight |
| Weather (offshore) | Seasonal - 5-15% | Variable | Schedule wells outside storm season |
4. Machine Learning and Digital Tools - Where They Add Real Value
4.1 What ML Actually Improves
Machine learning for drilling time prediction is effective in specific, well-defined use cases - and ineffective or misleading in others. Understanding the difference prevents both over-reliance and under-utilization:
| Use Case | ML Value | Requirement |
|---|---|---|
| ROP optimization in real time | High - identifies WOB/RPM sweet spots faster than human analysis | Minimum 50+ wells of drilling data in same formation |
| Bit wear prediction | High - predicts optimal pull point before performance cliff | Dull grade data correlated with surface drilling mechanics |
| Stuck pipe early warning | High - detects torque/drag trend changes 2-4 hours before event | High-frequency drilling data (1-second intervals minimum) |
| Well planning time estimates (frontier) | Low - insufficient training data, high geological uncertainty | Not recommended - use analytical methods instead |
| AFE estimation for development wells | Moderate - useful with 20+ wells in same field | Standardized data format across all offset wells |
4.2 Practical Digital Tools Available Today
- Landmark WellPlan / Halliburton PERFORM: Torque-drag and hydraulics simulation with time-depth prediction based on BHA design and mud properties
- Pason Drilling Intelligence: Real-time drilling efficiency monitoring with automated parameter optimization recommendations
- NOV NOVOS: Closed-loop drilling automation that continuously optimizes WOB and RPM to maximize ROP against a target - typically improves ROP 10-25% vs manual drilling
- Dataiku / custom Python models: For operators with large proprietary datasets - cluster analysis of offset well performance to generate formation-specific ROP distributions
5. Field Case Study - Deviated Well Drilling Time Estimate
Well profile: 12,500 ft MD, 65° max inclination, 4 hole sections, mixed carbonate and sandstone formations, 5 offset wells available.
Step 1 - Build normalized offset well ROP database by section:
| Section | Interval (ft) | P50 ROP (ft/hr) | P90 ROP (ft/hr) | Formation |
|---|---|---|---|---|
| 26" conductor | 200 | 180 | 120 | Unconsolidated |
| 17.5" surface | 2,800 | 95 | 65 | Sandstone/shale |
| 12.25" intermediate | 5,500 | 58 | 38 | Carbonate (hard stringers) |
| 8.5" production | 4,000 | 42 | 28 | Tight carbonate reservoir |
Step 2 - Calculate total well time (P50 estimate):
| Activity | Days (P50) |
|---|---|
| Drilling all sections | 14.2 |
| Tripping (6 BHA runs) | 4.8 |
| Casing and cementing (3 strings) | 3.5 |
| Logging and testing | 2.2 |
| BHA changes and miscellaneous | 1.5 |
| NPT allowance (12% of productive time) | 3.1 |
| TOTAL - P50 estimate | 29.3 days |
| P90 estimate (using P90 ROP + 18% NPT) | 41.8 days |
AFE basis: Use P50 (29.3 days) as the base estimate for planning purposes. Budget contingency = (P90 - P50) x daily rig rate = 12.5 days x $85,000/day = $1.06M contingency reserve. This is the quantified risk budget, not a vague "10% contingency."
Conclusion
Drilling time estimation done properly is a quantitative engineering exercise, not a judgment call. The difference between a P50 estimate and a P90 estimate for the same well is not uncertainty about whether the geology will be difficult - it is a quantified probability distribution built from normalized offset well data. The operators who consistently deliver wells on or under AFE are not luckier than their peers. They build activity-based time estimates, normalize offset well data before using it, quantify NPT by category rather than applying a blanket percentage, and use P90 rather than P50 as their AFE basis for high-risk wells.
The best investment of time before any well is building a proper time-depth database from offset wells and conducting an honest post-mortem on what drove NPT in each previous well. That institutional knowledge, systematically captured and applied, is worth more than any machine learning model trained on poorly labeled drilling data.
Want to access our drilling time estimation spreadsheet with activity-based breakdown and P10/P50/P90 analysis, or discuss a specific well planning challenge? Join our Telegram group for well planning discussions, or visit our YouTube channel for step-by-step tutorials on drilling time estimation and AFE preparation.
.jpg)
0 Comments