Call center performance is among the easiest problem-type to solve and optimize, because of the extensive volume of “reps.” (or agents etc.) in one or several office locations, and/or working at home or via calls handled by vendors.

Typical objectives/metrics are: sales, service (especially AHT and % first-time call resolution), and retention. (Call center turnover can exceed 100% and the solutions are found in hiring/on-boarding) to limit attrition/termination.)

Increasing collections (by aging categories), reducing bad debt, in businesses generally including healthcare, hospitals, cards etc.

Optimize marketing/advertising media mix, content, spend allocation and determine attribution.

Optimize sales channel performance without cross-channel cannibalization (feet-on-the-street, call center, stores/agents, on-line (by device type).

We offer seminars & workshops for all levels within an organization, from the data & analytics experts, senior managers & leadership, and the operations levels.

Increase mathematics test scores (and thereby, grades over semesters) in high-schools with hundreds of students (age unrestricted but most Nobi experience is ages 11 through 18)

Predictive models for academic achievement. Used for selection, coaching, gap closure. Grade school through university.

Increase % professional exam pass % (professional, trades, etc.)

Increase applicants through matriculation through all channels (awareness marketing, e-mail, social/digital, events, interviews/on-site visits, starts).

Comparative effectiveness research/studies for disadvantaged students.

Analytics/research to attain particular portfolio objectives.

Increasing sales volume (new and/or existing).

Forecasting key business metrics (e.g. revenues) broken out by division, product, territory etc. Use as feedback-control system to maximize sales, retention and identify/close performance gaps. All business types and sizes from a few $K (start-ups) to several $B. Addition of statistical controls to signal “spikes” or the onset of sales shifts in real time, for immediate solution.

CUSUM graphs amplify the sales signals and trace onset of shifts to the exact day (or equivalent, depending on data frequency such as hourly or monthly).

Noise-reduction to improve forecast accuracy. Over dozens of forecasts in many years, Nobi is typically attaining 1-2% annual forecast accuracy and in the region of 10-20% monthly.

A major feature of Nobi’s forecasts is the use of a mathematical device (called stationarity) which allows forecast to perform more accurately.

Optimize franchise and corporate sales (e.g. retail stores, travel agents).

Care Management (CM)

Reduce acute hospitalizations in chronic Medicare population of 5,000+ by 17% ($12.3M/year) by improved care and medications management.

The same approach was used in several other plans including one SPMI and special needs plans.

Predictive models to stratify patients into care categories from minimal to aggressive. Increased the CM yield overall, in utilization and HEDIS.

Study design and analysis to determine effectiveness of proposed new CM system. Findings were that the system was not yet viable, but we stumbled into a finding that home consultations by a particular type of clinical expert did augment the clinical model being used.

Optimize health survey, contact plan and accept rate for physician/NP consultations: Medicare/Medicaid and commercial plans. Outcomes include clinical and utilization. Approach duplicated many times since first application, developed in 2012.

Pain Management (including reduced % dependencies on selected medications).

Commercial wellness program: to increase engagement among 100,000+ employees with clinical pre-post metrics on outcomes.

Reduce re-admits via a hospital group with several units, by optimizing analytically the post-discharge transition and its critical time-windows.

Disease Management (DM)

Reduce acute hospitalizations in Medicare populations by improved care, in over a dozen distinct disease categories including diabetes, COPD, chronic heart disease, depression etc. with palliative care for some specific cohorts.

Autism episodes are disruptive in schools; this work designed and analyzed clinical studies to quantify the effectiveness of a new treatment program. Outcomes were strong with reduction in 5 types of disruption measured at 2/3 to 4/5 in 3 independent studies in pre-kindergarten through Grade 9 classrooms.

Medical Diagnostic Devices

Important metrics used in diagnostics include sensitivity (% true positive diagnosed correctly) and specificity (% true negative diagnosed correctly). This analytical work increased both, with almost all the gain in operator procedure/standardization. The hardest part is aligning a large population of operators/technicians/readers which is accomplished with the dynamic feedback-control, used in a novel way at each workstation.

Claims

Some STARS metrics required by regulation monitor claims processes. These can curtail sales if non-compliant and/or poorly performing against the industry cohort of comparable plans. This effort used simultaneous statistical design and control in 5 different processes at the same time to move from 98th. to 75th. percentile, suddenly then sustained.

Measurement error assessment and reduction in metrics used as above.

Workshop/Seminars

Under Nobi’s knowledge transfer program, our standard seminars were adapted in 2013 for hands-on work by MHAs seeking to add expert data science to their toolkit. Offered over varying durations from 1 to 5 days, all levels (basic to PhD.+ and diverse experience levels). Special module in statistical design and control for epidemiologists, statisticians, economists.

Propensity Scoring Models

Experimental validation/comparison of models used to assess CM/DM effectiveness. Use of operating-characteristic (O-C) curves to compare alternative models and optimize.

Provider Collections

AHT reductions, with collections and IB first-call resolution up.

Increasing military veteran hires.

Design and tune TSRU-based compensation systems for optimal feedback-control. Metrics include: operational performance and minimum volatility.

Sales training performance: establishing causality of training elements to sales.

Marketing/Media Mix & Spend

Multi-channel optimization of annual budget allocation to media and all channels: TV, radio, social, digital, e-mail, text, SEM/SEO, ROP, DM, etc. Several cases are run each year: in general there tends to be a 5% to 10% cost savings/re-allocation and net close rates increase in varying degrees by channel. This also solves for attribution.

Call Center Sales (IB, OB)

Many applications each year, increasing close-rate and lifetime value: fairly typical ROI in the $5M to $10M range for first year (for %-points on close rate). Subsequent years refine.

Warm Transfers

Call transferred live from vendors/partners – similar approach.

Paid Calls

Vendors selling lists are optimized both at list member selection and in closing. Uses predictive models but also orthogonal statistical designs to out-game vendor predictive algorithms.

Agent Sales

Typically 3rd. party vendors with one or a few under each ownership. Approach is the same as for retail store sales, but with less static/display and more script/people variables. Since agents have to participate voluntarily, statistical design has been most productive.

Incentives

In (especially) 3rd. party insurance agent sales, the effect of incentives is maximized by finding the variables agents respond to most. Variables found important here include agent selection and depth, cadence of incentives, frequency, amounts, interim feedback reports. Yield has been strong with the facility to as much as double the return (in policies written).

Claims

Similar to reducing claims expense in healthcare, but with variables involving adjustors, IB claims reporting centers, and allocation of claims types to specific specialty groups, also in changing the claims cycle times. An especially attractive feature has been reduction of attorney representation by 15 to 20%, saving money for claims and the insured.

Yield

Reduce defectives and rework in continuous or batch production.

Capacity

Increase capacity to deliver more product with the same head-count.

Design & Re-Design

In parts manufacture, diagnosis and solution to intermittent field failures in mechanical/electrical/electronic parts, by isolating the design and manufacturing variables that cause the failures. Uses a wider variety of statistical, mathematical and scientific tools such as finite elements, discriminant analysis and orthogonal experimentation.

Software Design

As above in software design for fewest “bugs” at release.

New Product Design

Market surveys and orthogonal conjoint analysis to find the optimal design and packaging, also price points and demand curve.

The standard marketing research and analytics are readily enhanced (or transformed) in their performance by powerful, modern analytics. Examples include causality/attribution (i.e. whether a given marketing element, program, campaign or event helped and by how much). In particular, powerful methods such as the lesser known CUSUM graphs and CUSCORE statistics allow particular marketing events to be separated from everything else going on in management, the market and the environment.

Conjoint analyses are considerably strengthened by adding orthogonal statistical conjoint design to find: not just if the proposed product will find market, but how to maximize it and what price points the market will bear.

Media Mix & Content

Optimizing advertising budget – spend allocation by channel, to increases sales at lower CPA.

Vacation Packages

Maximizing vacation packages sold at the right price. Updteam: leads and close %.

Channel Optimization

ROI for specific channels by finding the right content (e.g. TV, radio, direct mail, social-digital, ROP, inserts, FSI etc.)

All measurement systems (e.g. manufacturing, healthcare, clinical diagnostics, administrative, transaction/records) have error: none are perfect. Analytics easily determine if measurement is apt to do its job and if not how to improve simply. Key criteria are precision and accuracy.

Residential, commercial: predictive models and forecasting. Pricing models for given sales cycle time. Testing for sales

Store Sales “Comps”

Furniture store retail sales up 9.8% comps at +20% margin, in 3-months. The first year revealed the analytics could be done in a month. This was adopted. The initial work was sustained multi-year with additional increases layered on top each year (same analytical methods).

Implementing Test & Learn

Findings from test and learn cycles often fail to carry into implementation. There are specific, known reasons for this, having to do with how such retail tests are scaled. Here we used dynamic feedback-control to deliver the full test results and sustain.

Sales Forecasting

Forecasts 2 years ahead for a $2B business, broken out by a dozen business units and 15 to 20 main product lines. Accurate within 2% annually. Included seasonality.

With previously unpredictable sales patterns, more complex forecast analytics were used and found to forecast within about 10% monthly. These forecasts remained effective for some years, needing no updates, since they included a long-term growth predictor.

Sales Training

Finding causality: which training elements are effective in raising sales.

Sales Funnels

Sales funnels are simple feedback-control systems to which we add more sophisticated controls (mathematically designed but simple to use). % at each pipeline stage and goals are changed (up or sometimes down) to increase leads yield, conversion and close % (at less returns/cancels/churn). These are also adapted for atypical applications in healthcare, professional services etc.

Real Estate Sales

Sales forecasts including predictive models to optimize prices given “sell by” dates.

Predictive models for website visits/journeys

Analytics for progress through sales “funnels”

Capturing via “call and response” lead/prospect “footprints”

Self-learning algorithms to optimize content sent to targeted prospects and active accounts to maximize flow-though, close rates and retention/repeat customers

Increase enrolment and engagement in corporate wellness programs with causal shifts in outcomes (e.g. treatment shifts, meds., health outcomes, HEDIS)