One name in this week's screen has beaten consensus every quarter since late 2015. Another printed a streak of seventy quarters and was thrown out, because one of those quarters carried an estimate off by more than an order of magnitude. That contrast is the whole exercise. A beat streak is easy to compute and easy to get wrong, and the difference between a real one and an artifact lies in whether anyone checked the estimates that produced it.
Working from FMP's Earnings Surprises Bulk API across the S&P 500 and S&P MidCap 400, this screen counts a beat only where reported EPS came in strictly above consensus, treats a tie as a break, and then audits every quarter inside the surviving runs.
Key Takeaways
- Ties break streaks here, and that single rule removes more candidates than misses do: several long runs ended on a quarter where reported EPS matched consensus exactly.
- The longest streak in the screen, at 43 quarters, was not the longest the data produced. A 70-quarter run was excluded on estimate quality, which is the more instructive result.
- Four of the five are industrial or technology businesses with long order cycles, where the streak partly reflects management's visibility into its own book rather than operational surprise.
- A streak measures the distance between a company's guidance discipline and the Street's modelling, not the quality of the underlying business. Those are related but not the same thing.
Five Names That Keep Clearing the Bar
TE Connectivity (TEL)
Beat Streak: 43 quarters.
Next quarterly report: Nov. 4, 2026 — EPS: $3.07; Revenue: $5.27 billion (consensus).
TE Connectivity has cleared consensus every quarter since the run broke in October 2015, when reported EPS of $0.90 came in against a $0.93 estimate. Nearly eleven years without a miss is the longest verified streak in this screen, and the most recent quarter, $2.94 against $2.85, was typical of the pattern: a beat of a few percent rather than a dramatic one.
That consistency of magnitude is the part worth noticing. A company that beats by 3% quarter after quarter for a decade is not repeatedly surprising anyone. It is guiding to a number it is confident of clearing, and the Street is modelling close behind. For a connector and sensor business spread across automotive, industrial and communications end markets in most major economies, that implies an order book visible enough to forecast a quarter ahead with real precision, which is a genuine operational capability even if it does not read as one.
Where the streak becomes informative is in the mix underneath it. FMP's Revenue Geographic Segments API breaks the top line down by region, which matters for a company whose end markets move on different cycles in different places. Automotive content per vehicle, industrial order rates, and the currency effect on a heavily non-dollar revenue base are the variables that would eventually put pressure on the run.
RTX Corporation (RTX)
Beat Streak: 38 quarters.
Next quarterly report: Oct. 20, 2026 — EPS: $1.75; Revenue: $24.01 billion (consensus).
RTX has beaten consensus for 38 consecutive quarters, with the run ending in January 2017 on a tie rather than a miss: reported EPS of $1.56 against an estimate of $1.56. Under the rule used here that breaks the streak, which is worth stating plainly because a looser definition would extend it considerably. The most recent quarter came in at $1.89 against $1.66, a wider margin than the pattern usually produces.
One structural caveat belongs with this name. The streak spans the 2020 combination of United Technologies and Raytheon, so the earlier quarters in the run belong to a differently constituted company. The surprise history is continuous in the data because the reporting entity is continuous, but a reader should treat the pre-2020 portion as describing a predecessor. That does not invalidate the streak; it changes what the streak is a statement about.
The more useful comparison is between what the company guides and what the Street carries, and FMP's Financial Estimates API holds those forward lines by period. Aftermarket volumes in commercial aerospace, defence backlog conversion, and supply chain throughput are the operating variables that sit behind a beat pattern this long, and any of them loosening would show up in the estimate revisions before it showed up in a miss.
Marsh & McLennan (MRSH)
Beat Streak: 32 quarters.
Next quarterly report: Oct. 15, 2026 — EPS: $1.98; Revenue: $6.66 billion (consensus).
Eight years of consecutive beats, broken last in July 2018 by a single cent, $1.10 reported against $1.11 expected. The June quarter came in at $2.96 against $2.88. For an insurance brokerage and consulting business, this kind of run reflects a revenue model built on recurring commissions and fees that renew rather than being won again each period, which makes the forward quarter unusually forecastable from the inside.
That said, the mechanism deserves scepticism rather than admiration. When a business can see most of its next quarter's revenue before the quarter starts, beating consensus by one or two cents is a choice about where to set expectations as much as an operational outcome. The interesting question is not whether the streak continues but whether margins are still expanding underneath it, because a fee business can hold its beat rate while its economics slowly flatten.
That is a ratio question rather than an earnings question, and FMP's Financial Ratios API carries the operating and net margin series across the full period of the streak. Organic revenue growth by segment, the consulting versus risk and insurance services split, and margin progression are the lines that distinguish a durable model from a well-managed expectations process.
Lowe's Companies (LOW)
Beat Streak: 29 quarters.
Next quarterly report: Nov. 18, 2026 — EPS: $2.88; Revenue: $22.41 billion (consensus).
Lowe's has not missed since May 2019, when $1.22 landed against a $1.33 estimate, and the August quarter continued the pattern at $4.40 against $4.22. What makes this the most surprising entry in the screen is the backdrop. The streak runs straight through the pandemic home improvement boom, the subsequent demand normalisation, and a prolonged stretch of elevated mortgage rates suppressing housing turnover, which is the single most important external driver of the category.
Holding a beat record through that sequence says something specific: the cost structure and the expectations process adjusted faster than the demand environment moved. Comparable sales have been negative for extended periods during this run while EPS still cleared consensus, which points to buybacks, margin management and guidance discipline doing work that revenue was not. That is a legitimate form of execution, and it is also a reminder that a beat streak is compatible with a shrinking top line.
Separating the two requires normalised metrics rather than headline figures, and FMP's Key Metrics TTM API provides the trailing return, margin and per-share series that show how much of the earnings line is being produced by share count. Comparable sales, the professional customer mix, gross margin, and the pace of repurchase are what determine whether the next print is another routine clearance.
Pfizer (PFE)
Beat Streak: 22 quarters.
Next quarterly report: Nov. 3, 2026 — EPS: $0.79; Revenue: $15.93 billion (consensus).
Pfizer's run has been intact since February 2021, when $0.42 came in against $0.46, and the most recent quarter delivered $0.77 against $0.682. The composition of that streak has changed almost completely across its life. Its early quarters were produced by pandemic-era vaccine and antiviral revenue at a scale no model had a precedent for. Its recent quarters have been produced by a much smaller, more conventional revenue base with cost reduction programmes doing a substantial part of the work.
That transformation is exactly why this streak is the most interesting one in the screen and the least comparable to the others. TE Connectivity and Marsh & McLennan beat consensus because their businesses are predictable. Pfizer beat consensus through a period when its business was anything but, which means the mechanism has shifted from forecasting accuracy to expectations being set conservatively against a declining base. Those two things look identical in a surprise history and are completely different in what they imply.
Cash generation is where the distinction is visible, and FMP's Cash Flow Statement API shows how much of the reported earnings is converting, and how capital allocation has changed as the revenue base normalised. Patent expiry exposure across the back half of the decade, pipeline contribution, and the realised savings from the cost programmes are what will determine whether the beat pattern outlasts the cost lever.
What a Long Streak Does and Does Not Prove
The most useful output of this week's screen was a rejection. One company recorded seventy consecutive beats, which would have been by far the longest run in the universe, and it did not survive an audit of the quarters inside it. One period carried a consensus estimate roughly an order of magnitude below the reported figure, which is the signature of a stock-split-unadjusted estimate rather than a forecasting failure. Two other quarters inside the run counted as beats only because reported losses were smaller than expected. Mechanically a beat, analytically not the same event. A screen that ranks by streak length without opening the quarters will surface that name first every time.
Definition does comparable work. Treating a tie as a break removed several long runs outright, including one that would otherwise have topped this list. That choice is defensible and it is not obviously correct, but it has to be made explicitly and applied consistently, because a streak computed under one rule is not comparable to a streak computed under another. Reporting the exact date and values of the breaking quarter, as this screen does for all five names, is what makes the number auditable rather than asserted.
Once the counting is sound, the harder question is what the streak measures. The five names here cluster into two mechanisms. TE Connectivity, RTX and Marsh & McLennan operate with long order books or recurring revenue, where visibility into the coming quarter is genuinely high and the streak reflects that visibility. Lowe's and Pfizer ran their streaks through environments where visibility was poor, which means the consistency came from cost management and conservative guidance rather than from forecasting. Both produce the same number. Only one of them is evidence about the business.
Testing which is which takes several datasets in sequence. The Earnings Report API supplies the full per-company surprise history needed to measure the size of each beat rather than just its existence, and a streak of consistent 2% beats reads very differently from one with wild dispersion. The Financial Estimates API shows whether analysts have been revising toward the company over time, which is what a guidance-driven streak looks like from the outside. The Stock Split Details API is the direct check on the failure mode that disqualified the seventy-quarter name, since split-unadjusted consensus is the most common corruption in any surprise dataset. The Earnings Transcript API carries the guidance language itself, which is where the distinction between conservative framing and genuine visibility is usually audible before it is measurable.
The fundamentals then have to corroborate all of it. The Cash Flow Statement API and the Key Metrics API establish whether earnings consistency is accompanied by cash conversion and stable returns, or whether share count and cost programmes are carrying the line. The Historical Stock Grades API adds how analysts have responded across the same period, and a long beat streak paired with steadily falling ratings is a meaningful contradiction rather than an anomaly. Holding surprise history, estimates, statements, corporate actions and transcripts in one place is what turns a count into an argument, which is the practical case for running this across the FMP platform rather than assembling it from separate sources where the split adjustments never quite agree.
None of these five is predictable because it has a streak. Each one is worth examining because the streak raises a specific, answerable question about how the business and its guidance interact, and because the next print, in every case within the next two months, supplies the answer.
Building a Repeatability Screen with FMP Data
Screening for consistent earnings outperformance works best when you start from the entire dataset rather than a curated list of names. Pre-selecting companies introduces bias upfront—especially when the objective is to uncover patterns that may not be obvious within familiar coverage. A cleaner approach is to begin with the full distribution of earnings outcomes and let the filtering process surface repeat performers on its own.
The FMP Earnings Surprises Bulk API is the practical entry point for that process. It standardizes quarterly EPS results against consensus estimates across a broad equity universe, which makes it possible to scan for surprise patterns without needing to stitch together multiple sources. Before running any calls, the only requirement is a valid API key.
1. Pull Bulk Earnings Surprises
Begin by hitting the Earnings Surprises Bulk API, which aggregates every quarterly EPS surprise — positive or negative — for the year you specify:
|
https://financialmodelingprep.com/stable/earnings-surprises-bulk?year=2025&apikey=YOUR_API_KEY |
Sample Response:
|
[ { "symbol": "AMKYF", "date": "2025-07-09", "epsActual": 0.3631, "epsEstimated": 0.3615, "lastUpdated": "2025-07-09" } ] |
From here, the first cut is mechanical: isolate the entries where epsActual > epsEstimated. That gives you the universe of names that beat expectations at least once during the period — essentially a raw pool before you evaluate whether any of them can deliver that result consistently.
2. Retrieve Company-Level Details
With that universe in hand, the analysis moves from identifying events to evaluating consistency. For each ticker that cleared the first filter, pull its complete quarterly earnings history using the Earnings Report API:
|
https://financialmodelingprep.com/stable/earnings?symbol=AAPL&apikey=YOUR_API_KEY |
Looking at the complete sequence of reported quarters makes it possible to evaluate frequency and clustering. This is where judgment enters the workflow. Some analysts require three or more consecutive beats to qualify as a streak; others impose minimum surprise thresholds or remove near-zero deviations. The parameters can be adjusted, but the intent stays the same: separate sustained execution from statistical noise.
By this point, the screen has moved beyond identifying isolated surprises. What began as a broad event scan turns into a structured assessment of earnings reliability, highlighting companies where internal forecasting and operational control have proven more consistent than the market's expectations over time.
Scaling the Screen: Testing the Signal Across Broader Coverage
A repeatability screen earns its credibility not by how many companies it includes, but by how well it holds up as the data environment becomes less predictable. The most effective way to test that is to widen the scope in stages—starting with tightly modeled segments and moving outward into areas where estimates are looser and coverage is thinner. The goal is simple: see if the signal survives as conditions become less controlled.
The first pass sits within the Free plan, where the dataset is anchored in widely covered large-cap names like Apple Inc., Alphabet Inc., and JPMorgan Chase & Co.. This is the most efficiently modeled part of the market—analyst coverage is dense, estimates tend to cluster, and expectations are generally well-defined. If a streak-based approach produces consistent outputs here, it suggests the framework is functioning on stable ground rather than reacting to noise in the data.
Moving into the Starter plan expands the universe to include a wider range of U.S. equities, bringing in smaller-cap and more specialized companies. At this level, estimate dispersion becomes more visible as coverage thins out. That shift acts as a natural pressure test. When the same criteria continue to surface repeat performers, the pattern begins to look less dependent on tightly managed forecasts and more tied to actual operating consistency.
The final step, via the Premium plan, extends the screen across international markets such as the U.K. and Canada. The methodology remains unchanged, but the inputs become more varied. Differences in reporting conventions, sector mix, and estimate reliability introduce additional complexity. Applying the same framework across these regions helps validate that the signal—earnings consistency—remains comparable even as the underlying data becomes more heterogeneous.
From Analyst Workflow to Standardized Research Framework
The transition from individual screening model to firm-wide research infrastructure usually begins quietly. An analyst builds a process to track recurring earnings surprises, another team starts referencing the outputs, and over time the methodology becomes embedded in broader coverage discussions. What starts as a desk-level workflow gradually turns into a standardized framework for evaluating operational consistency across sectors and market regimes.
That evolution is rarely driven by top-down mandates alone. In practice, the strongest push toward standardization tends to come from analysts closest to the data — the people refining filters after each earnings cycle, stress-testing assumptions against new quarters, and identifying where consensus models repeatedly fail to adjust fast enough. Once a workflow demonstrates that it can consistently surface meaningful signals, the inefficiencies of fragmented research processes become harder to ignore.
Most firms already recognize the symptoms: multiple teams maintaining slightly different versions of the same screen, inconsistent threshold definitions across sectors, isolated spreadsheets with limited transparency, and duplicated effort spent reconciling conflicting outputs. Over time, those inconsistencies become operational friction. They slow collaboration, complicate audit trails, and make it difficult to determine whether differences in conclusions come from actual analytical disagreement or simply from variations in methodology.
A standardized framework changes the conversation from “who built the model” to “how the organization defines the signal.” Shared dashboards replace disconnected files. Screening criteria become reviewable rather than buried inside private workflows. Inputs, revisions, and assumptions can be tracked across teams, creating clearer governance around how data-driven conclusions are formed. For research groups operating across multiple sectors or geographies, that consistency becomes increasingly important as workflows expand beyond the analysts who originally designed them.
At that stage, the challenge is no longer discovering the signal — it is preserving methodological consistency as adoption widens across the organization. Centralized infrastructure such as the FMP Enterprise plan becomes relevant not because it changes the analytical framework itself, but because it helps maintain unified data access, version control, and cross-team visibility once the workflow moves beyond a single desk. The objective is continuity: ensuring that the same earnings-consistency signal is being measured the same way across the broader research platform.
Carrying the Streak Into the Next Print
All five of these runs face a test inside the next two months, which is the useful thing about a screen built on scheduled events rather than sentiment. Rebuilt weekly from the Earnings Surprises Bulk API, the list updates itself as those prints land, and the breaks will be as informative as the runs.
Want more? Explore our earlier article: Weekly Signals Desk | Concentrated Analyst Revisions via the FMP API (Sept 7-11)
Disclosure: Signals Desk content is provided for informational and analytical purposes only and does not constitute investment advice or trade recommendations. The analysis reflects interpretation of market data and publicly disclosed or third-party information, including data accessed via Financial Modeling Prep APIs, at the time of publication. Signals discussed are probabilistic, can be wrong, and may change as market conditions and consensus data evolve. This content should be considered alongside broader research, individual objectives, and risk assessment.
