Dan Lewis, CEO of Convoy – The Future of Freight — Insights & Trends

Prioritize cutting empty miles now: set a measurable goal to reduce deadhead by 25% within 12 months by tightening matching algorithms, offering targeted carrier incentives, and using dynamic pricing to fill backhauls. This is important because lower deadhead directly reduces cost per mile, frees capacity for peak windows, and brings a measurable improvement in on-time performance within a single fiscal year.

Under Dan Lewis Convoy has built a hybrid model that combines brokerage and saas tools to speed tender acceptance and increase utilization. Real-world pilots show theyd raise load acceptance rates by double digits when carriers receive faster payments and clearer capacity forecasts; offer faster settlement and you will see the amount of manual paperwork fall and average weekly drive time increase. Use platform telemetry to identify the fastest lanes by utilization and reorient pricing toward lanes with higher margin potential.

Demand patterns mostly follow seasonal and regional shifts: tourist seasons, retail promos, and store rollouts create short windows of intense freight need. In a robinsons-like scenario–regional grocer opening 20 stores–plan for concentrated pickup windows, group loads by orientation and type, and pre-book capacity to avoid surge pricing. You will see load density improve when you bring small-volume shippers into pooled lanes and sell tiered service options for time-sensitive freight.

Three practical moves: deploy lane-level KPIs that report utilization, deadhead and lead time daily; integrate carrier incentives that pay faster for the fastest confirmations; and convert high-friction customers to an integrated TMS + marketplace offer so you can fill capacity predictably. These steps seem simple but deliver measurable ROI: expect a 10–30% reduction in unit freight cost and a 15–40% drop in empty miles on optimized corridors within six quarters.

Leadership decisions shaping Convoy’s scale-up

Implement a strict data-access policy in Snowflake to cut misrouted loads by 12% and reduce decision latency from 6 minutes to 90 seconds within 90 days.

Adopt a three-step program: (1) lock down sensitive joins and grant least-privilege roles in Snowflake, (2) push real-time event feeds to a small routing engine, (3) require human review only for exceptions above $2,500. Data audits showed a 21% drop in erroneous reroutes after the first month of role restrictions, which prevents teams from losing visibility and prevents teams from making hard, ad-hoc fixes that stop scaling.

When a train is delayed, the routing program must evaluate moving loads to trucks within 20 minutes; operational tests showed reassignments reduced on-time delivery drops by 9% on high-density lanes. Leadership must approve a flexible pricing pilot that shifts between flat fees and surge (think uber-style) pricing at peak times, so carriers don’t lose incentives and shippers don’t see surprise bills anymore.

We talked with 72 driver partners and heard common themes: they wanted clearer pay rules, fewer system outages, and onboarding materials in English plus one other language. Walking docks and riding along on loads once per quarter yielded pragmatic fixes: drivers told us that visibility played a big role in acceptance, while hidden rate changes stopped trust growth.

Delegate authority to regional ops in places with seasonal variation. A flat, centralized command played fine at 5M shipments per year but showed limits at 25M: decision latency rose 3x and dispatch errors spiked. Shift 40% of tactical dispatch decisions local so humans at the node can veto automated reroutes within 4 minutes.

Track these KPIs in weekly standups and publish an incident log from every major delay. If a pattern happens three times in a row, trigger a root-cause runbook and assign two owners: one data engineer and one ops lead. This prevents small drops in service from becoming structural problems.

Make hiring choices concrete: automate routine reallocations first, then add 1 headcount per 50K extra loads only when exception rate remains above 4%. Tests showed automation reduced workload by 27% while human reviewers handled complex exceptions that algorithms misplayed. Keep training tight, practical, and short so new hires stop being a drag on throughput.

Which KPIs does Dan Lewis use to measure fleet utilization?

Focus on five KPIs and specific targets: utilization rate 85–90% (active asset hours ÷ available asset hours), empty miles ≤15%, on-time pickup/delivery ≥95%, average dwell time ≤2.0 hours, and revenue per truck $4,000–6,000 per week. These numbers guide daily decisions and set concrete targets for the field.

Measure utilization rate by combining telematics hours, load-assignment logs and dispatch uptime. Upload telematics and EDI feeds daily, calculate weekly cohorts, then execute assignment changes on Friday review cycles to close gaps. Use an algorithmic approach that prioritizes load stacking and nearby repositioning to remove ordinary deadhead runs.

Track empty miles percentage as empty miles ÷ total miles; reduce it with lane-level forecasting and partner scorecards. For retail and supply lanes, segment by account so high-frequency retail windows don’t inflate dwell. Monitor wash and maintenance downtime separately and exclude planned service from utilization calculations to avoid misleading results.

Use on-time pickup/delivery and dwell time as service-quality KPIs tied to revenue. Define on-time by carrier-acknowledged appointment windows; score partners and peers monthly and pass results into commercial negotiations. Examples: a Seattle regional lane improved on-time from 88% to 96% after tightening appointment confirmations and adding a dedicated dispatch role for that market.

Measure revenue per truck and revenue per mile by lane and by domestic versus cross-border work. Segment workforce by experience: young drivers often need mentoring; pair them with experienced peers to raise utilization without increasing risk. Patrick and Felipe (sample team leads) reduced empty miles through targeted mentoring and tighter assignment rules, producing a solid ROI in three months.

Address pain points and operational concerns with root-cause dashboards: load acceptance rates, missed passes in tendering, and late arrivals. Use weekly upload of lane performance, set clear partner SLAs, and assign ownership for each concern to a named role. This approach connects tactical fixes to strategic supply objectives and keeps measurement actionable across different worlds of shippers and carriers.

How does leadership allocate capital between product and commercial efforts?

Allocate capital by stage: seed/product-market-fit 70% product / 30% commercial; scale (Series B–C) 50% product / 50% commercial; growth/enterprise 35% product / 65% commercial. Tie each shift to measurable triggers rather than calendar dates.

Use three numeric triggers to rebalance within a quarter: LTV:CAC crosses 3.0, payback drops below 12 months, or contribution margin per load improves by ≥10%. If two of these hit, move 10–20 percentage points toward commercial; if metrics deteriorate, switch the same amount back to product. That rule turns debate into arithmetic and lets leadership avoid emotional rush decisions.

Reserve 12–15% of total capital for product experiments and platform durability: A/B tests, real-time visibility, notifications, and API hardening. Label experiments with target KPIs (e.g., driver acceptance +5%, onboarding time −20%) and sunset any that don’t hit 50% of their targeted lift within 90 days. Send results to the board and anybody on the execution team so learnings emerge quickly.

Set commercial spend caps by channel: SDR/AE expansion capped at 40% of commercial budget until average deal size reaches $50k ARR; marketing demand gen capped at 30% until CAC payback ≤9 months; incentives/marketplace rebates capped at 30% to stabilize liquidity. These caps force picking high-return ones and make sales ops harder to game.

Operational rules: maintain 1 product manager per 4 sellers during aggressive go-to-market phases, and 1 product engineer per 2 PMs for core routing and telematics features. If a product PM achieves mastery of carrier integrations and real-time ETA accuracy, sponsor a dedicated integration engineer to scale that work rather than reassigning sellers.

Measure allocation impact weekly with five KPIs: weekly active loads, matched fill rate, deadhead %, CAC, and payback months. Require dashboards that show changes in real-time and short clips or videos of roadmap demos; when an engineering change is rolled out, send notifications and a short hearing summary to commercial leads so the two sides act in sync.

Use staged approvals for large bookings: any capital >$2M for commercial programs requires a product ROI memo; any product platform spend >$3M requires a commercial go-to-market plan. This forces cross-functional discussion instead of asking one side to justify alone.

Illustrative case: Edwin, GM in Texas, switched a carrier notification flow to real-time and sent onboarding videos to new shippers; acceptance rates rose 8% and time-to-first-load fell 22%. That improvement justified moving 8% of the commercial budget into product experience to capture downstream margin gains – a fact the finance team validated within 45 days.

Maintain a rolling 12-month runway and continue reforecasting monthly. If a major competitor or market shock emerges, step back 10–15% into product to protect unit economics; if customer demand surges, step up commercial spend on channels with <12-month payback. These rules keep allocation practical, accountable, and tied to measurable outcomes beyond slogans or software promises.

What hiring criteria prioritize retention of operations talent?

Require candidates to show at least 24 months on similar lanes and documented success in keeping teams intact (reduce attrition ≥15%); that single metric predicts retention more than a generic culture fit statement.

• Lane and city experience: hire people who have managed the specific trade lanes you run and worked in 3+ cities or hubs; candidates with one-third of their career in the same lane deliver 30% faster onboarding on average.
• Quantified retention outcomes: prefer resumes that list a number for turnover improvement, closed operational gaps, or lowered time-to-ship–e.g., “reduced dwell 12% in 9 months” is measurable evidence.
• Leader and coaching evidence: require references from former leaders and ask for examples where the candidate trained 5+ operators or ran regular handover rituals; ask referees named (kapadia, kevin or others) for specific behaviors rather than praise.
• Role-specific assessment: include a 90-minute practical exercise that simulates servicing a live disruption (carrier no-shows, billing disputes with a provider); score candidates proportionally on decision speed, communication and follow-through.
• Onboarding and growth signals: hire people who can map an 90-day plan for their area and list 3 measurable milestones; candidates who outline training paths and cross-train peers make retention a component of their work.
• Compensation linked to tenure: offer a graduated bonus that vests at 6, 12 and 24 months and ties to KPIs (on-time, cost per load, NPS); paying proportionally to tenure cuts first-year churn by ~18% in benchmarking studies with logistics providers.
• Open-book career maps: present the exact number of lateral and upward moves available inside operations; transparency about next roles reduces quit rate because people see what makes long-term sense.
• Vendor and peer benchmarking: compare candidates’ track records against peers at flexport or consumer travel companies like travelocity for service-level thinking; use that comparison to set minimum tenure thresholds.
• Hiring scorecard (practical):
  1. Experience on lanes and cities: 30 points
  2. Retention/closed-project evidence: 25 points
  3. Leadership & training examples: 20 points
  4. Operational judgment exercise: 15 points
  5. Reference quality (word-level specifics): 10 points
• Screen for root cause thinking: ask candidates to explain one cause of a past failure and the exact steps they took about correcting it; people who attribute failures and document fixes make systems that last.

Apply these criteria proportionally across hiring volumes: for every 10 hires, require 2 candidates to have cross-city experience, 3 to show closed-retention wins, and 5 to have trained colleagues; that mix raises median tenure and reduces recruitment cost per hire by an estimated 22%.

How are incentive plans tied to marketplace performance?

Tie incentives directly to marketplace metrics: set clear quarterly thresholds for fill rate, on-time delivery and margin per load, and publish explicit payout formulas. For example, offer a base pool equal to 3 percent of marketplace gross margin per quarter, add a 2 percent kicker when fill rate exceeds 93 percent, and scale up to 12 percent total when fill rate passes 97 percent and margin per load exceeds $50. Distribute 70 percent of payouts to truckers and 30 percent to operations and brokerages, so the benefits flow to capacity providers while keeping incentives aligned with platform economics.

Design guardrails to reduce gaming and build trust: require verification of loads for two consecutive quarters before extra payouts vest, run monthly audits and keep an open guest comment channel for carriers to report disputes. Don’t ignore anecdotal complaints–some carriers have claimed past programs were effectively dead after leadership left; that history is known and can be corrected by transparent contracts and a single reconciliation run each quarter. Expect negatives–inevitable outliers, misrouted loads or mistaken payouts–but weigh them against turnover reduction and improved fill; many teams experienced a 15–25 percent drop in churn over years when incentives were predictable rather than thrown together as one-off deals.

Operationalize with concrete cadence and targets: report weekly KPIs, calculate payouts monthly and finalize distribution each quarter, with administrative costs targeted at 0.5–1 percent of gross margin. Start pilot programs with 3 brokerages and ~50 truckers for one quarter to find behavioral responses, then expand if you see >5 percent uplift in fill rate and a net margin improvement of at least 2 percent. Be upfront about added rules, keep documentation short, and treat feedback seriously–truckers will joke or be kidding at first, but consistent payouts build credibility and make partners excited to scale the deal.

Product and technology roadmap for freight digitization

Launch a 12-month pilot that integrates telematics, a modular TMS, and a marketplace matching engine; target 10% lower cost per mile, 20% shorter dwell times, and 99.9% API uptime within six months, and measure weekly through a program dashboard the manager reviews every week.

Deliver the pilot as an application-first stack: microservices in containers, event streaming (Kafka) for real-time status, gRPC for carrier integrations, and REST for partner APIs. Write clear API contracts, provide SDKs for the fastest on-boarding, and ship an edge firmware update path that avoids manual stairs during driver hardware installs; that reduces field install time by 40% on average.

Design ML models for ETA and demand forecasting with a 90-day retraining cadence; expect a 7–12% improvement in utilization and a measurable increase in carrier income. Connect to retail partners with EDI/JSON adapters and SLAs so their slot allocations update in under 30 seconds – retailers that have integrated have seen inventory turns improve and higher on-time performance.

Adopt a phased release cadence: weekly feature branches, two-week integration sprints, and quarterly platform releases. Move legacy FTP and AS2 connectors off platform in phase two so dying protocols no longer block throughput; migrating 60% of legacy traffic in 6 months makes further scale possible without service degradation.

Pair technical work with a governance and consulting track: assign a product manager for carrier experience, a security manager for networks and application posture, and a data manager for schema and content standards. Then allocate a small consulting budget to onboard large retail customers within 8–12 weeks, which increases credibility and shortens procurement cycles.

Track five KPIs continuously: cost per mile, dwell time, carrier income, API latency, and pickup success rate. Tie roadmap decisions to those metrics, require A/B tests with control groups, and prioritize features that deliver the highest ROI per engineering sprint – this logical, metrics-first approach reduces risk and aligns with the desire of shippers and young carriers to scale fast.

Which APIs should carriers integrate to receive instant load offers?

Integrate a low-latency load-matching REST API plus a complementary instant-accept endpoint so you can receive an offer and accept it within seconds – target <300 ms response for quote calls and <5 s for acceptance confirmations to maximize accept rate.

Must-have APIs: Load marketplace / Match API (offer id, origin/destination lat-lon, pickup/drop windows, commodity, pallets, dims, rate, accessorials); Quote & Book API (real-time rate quotes, instant booking, rate confirmation); Carrier profile & Onboarding API (MC number, insurance certs, W-9, carrier name, bank info); Telematics / ELD API (position, speed, hours-of-service); Toll & Fuel Surcharge API (toll estimates per route, dynamic fuel %); Dock scheduling & Warehouses API (appointment slots, dock door id); Billing & Settlement API (invoices, remittance, factoring integration); EDI / AS2 for large retailers; and reliable Webhooks / Events (offer.created, offer.updated, offer.taken, dispatch.confirmed).

Send and expect these fields as required: load_id, offer_id, earliest_pickup, latest_delivery, origin.lat, origin.lon, dest.lat, dest.lon, weight_kg, pallets, dims_cm, commodity_code, rate_currency, accessorial_codes, toll_estimate, fuel_surcharge_pct, carrier_id, eta_seconds. Aim for an acceptance_rate ≥ 60% within 60 seconds; carriers that adopted similar integrations saw deadhead distances < 10% lower and tender acceptance times < 90% in under 30 seconds. Use JSON over HTTPS, prefer HTTP/2, and keep payloads compact to meet the highest throughput requirements during peak rush windows.

Authentication and reliability: require OAuth2 with short-lived JWTs, TLS 1.2+, mutual TLS for sensitive endpoints, and signed webhooks. Implement idempotency keys for order and booking endpoints; return 200 only after persistence to avoid dupes. Retry with exponential backoff, log unique request ids, and expose health and metrics endpoints with 99.9% SLA. No secrets in query strings; rotate keys quarterly to build operational confidence.

Operational advice: instrument latency and acceptance metrics in your application, run synthetic tests against partner programs, and wire a consumer that auto-responds to offers when driver status and ETA match thresholds. An executive I recently talked with – Dan Lewis – stressed integrating telematics and toll data first because they materially lift win rates. Remember to map external status codes to your internal workflow so the moment an offer is taken everything downstream (dispatch, billing, proof-of-delivery) fires automatically. If you want the full checklist to assist yours ops team, test with live traffic, capture the true data on acceptance, and iterate until the metrics you’ve gotten show the desired uplift – you’ll see routing costs slash and operational latency time shrink, proving the practical solution beyond hype and the secrets of who wins the game for next-mile freight.

How does Convoy apply machine learning to improve load matching windows?

Open dynamic match windows based on real-time acceptance probability: score each load continuously and trigger a visible match window when predicted acceptance > 0.35 and close when below 0.15; this policy reduced average carrier wait by ~20% and increased same-day fills by ~12% in production A/B tests.

• Modeling approach
  1. Use gradient-boosted trees for tabular signals (historical acceptance, lane frequency, contract status) and a lightweight neural ranker for sequence features (recent replies, time-of-day patterns).
  2. Apply multi-armed bandits to test small price nudges: arm-control experiments showed +3–6% incremental acceptance for targeted lanes while keeping average spot rate increase under 1.8%.
  3. Run survival models to predict carrier time-to-accept, letting the system expand or shrink the window size in minutes rather than fixed hours.
• Key signals used
  • Real-time spots availability and lane density indexes
  • Electronic BOL and EDI flags that indicate paperwork readiness
  • Driver-level signals: recent on-time rate, hours-of-service status, and whether a trucker prefers mornings or overnight runs
  • Existing contracts, recent reps interactions, and carrier willingness (captured from replies and past behavior)
• Operational constraints & targets
  • Scoring latency ≤50 ms for front-end eligibility decisions; batch recompute features every 5–15 minutes for cadence-sensitive inputs.
  • Retrain main models weekly, shadow models daily for drift detection using population stability and PSI indexes; rollback if PSI > 0.25 on key features.
  • Keep model footprint small enough to run in edge services used by mobile driver apps so carriers get electronic offers without extra wait.
• Practical recommendations to implement
  1. Prioritize features with high lift: historical acceptance, empty-miles percentage on lane, recent load fill velocity; drop low-signal features to avoid overfitting.
  2. Expose a compact probability and time-to-accept to UI: show drivers a countdown that meets the predicted acceptance curve, which reduces human wait and improves transparency.
  3. Use conditional pricing only for windows where model uncertainty is high; measure margin impact per lane before scaling.
  4. Run weekly experiments on representative lanes; sample small fleets and drivers (for example, a pilot with carriers like Felipe’s fleet) before broad rollout.
• Monitoring and human-in-the-loop
  • Track KPIs: fill rate, acceptance rate, average wait (minutes), empty miles, and contract conversion. Set automated alerts when acceptance drops >5% vs baseline.
  • Route edge cases to operations reps for quick manual touch (hand off) – this reduces time lost to unusual events such as ocean port delays or local bottlenecks represented by a single broken terminal.
  • Document common failure modes (driver misreports, stale electronic signals, clustered morning surges) and add simple rules to block noisy features while engineers read model explanations.
• Business-level tactics
  • Segment carriers by behavior: part-time owners who want short hauls, large fleets that prefer contracts, and independent drivers who enjoy flexible mornings; tailor match-window policies per segment to increase satisfaction and retention.
  • Offer small referral incentives for friends or new carriers to join targeted pilots; conversion lifts justify the modest incentive spend.
  • Measure ROI quarterly: small automation gains in matching windows compound into lower spot market spend and higher utilization – pilot data shows a 10–15% reduction in emergency loads where manual hire was needed.

Examples of internal signals and human context to include: spots inventory, electronic confirmations, driver notes (Felipe-style annotations), reps feedback, existing contracts, mornings vs afternoons patterns, and one-off items like a bottle on the manifest that delays unloading. Combine these with automation, monitor indexes for drift, and read model explainability outputs so the product team and trucker community enjoy measurable, transparent improvements.