Dial-by-Name Directory vs. Phone Menu for Small Teams

Executive Summary

A dial-by-name directory helps callers who know a person but lack that person's extension. A purpose-based phone menu helps callers who know the job they need done, such as billing or an appointment. These are compatible routing choices: Webex documents name dialing and extension dialing as distinct auto-attendant actions, rather than treating a directory as a replacement for the attendant. ^[1] For a small professional firm, caller knowledge should determine the design before employee count does.

The implementations differ materially. Talkroute documents manually created name-derived extensions with a consistent three- or four-digit length and a submenu for shared patterns. ^[2] Dialpad documents first-name keypad entry, office or company scope, and support activation for extension dialing; its extension-management page separately identifies Enterprise eligibility. ^{[3] [4]} Microsoft Teams allows partial-name searches but requires a complete extension, illustrating why the two tasks need separate tests. ^[5] A vendor's directory label alone does not establish the caller experience.

The central recommendation is to test a short purpose menu against a name route using the same realistic caller tasks. The National Institute of Standards and Technology (NIST) identifies task time, errors, and successful completion as usability measures. ^[6] This report supplies editable caller-knowledge, collision, and acceptance-test worksheets. Its fictional roster deliberately produces matching keypad prefixes; those results are calculations, not evidence about how often collisions occur in real firms. The keypad mapping comes from the International Telecommunication Union (ITU). (Source: www.itu.int) No independently measured speed advantage for either method was established by the reviewed sources.

A directory also creates a maintenance obligation. Microsoft's reference describes address-book updates taking up to 24 hours. ^[5] Its setup guide describes up to 36 hours for a new user's listing. ^[7] These describe different update contexts and should not become a guaranteed removal deadline. Keep a route for callers who cannot find a name: World Wide Web Consortium (W3C) supplemental guidance recommends simple error recovery with access to a person when errors persist. ^[8] Keep the menu when purpose routing passes the firm's tests, add a directory when named-person tasks justify it, retain selective direct numbers where useful, and use a human receptionist when the caller needs conversation.

Introduction and Background

A caller may request Alex, ask for accounts, or need help identifying who handles a problem. Those statements describe different information problems. For professional firms with 2 to 20 people in the United States and Canada, the useful question is whether letting callers spell a name resolves their problem more reliably than asking why they called. That is a testable local hypothesis, not a staffing threshold.

An auto attendant is the routing layer that offers choices and sends calls onward. A directory is a way to identify a destination within that flow. In Webex Calling, administrators can assign separate menu actions for name dialing and extension dialing. ^[1] This distinction matters when comparing “dial by name directory vs auto attendant”: the directory can be one of the attendant's options.

A public direct number is a number a caller can dial without first selecting someone through the shared menu. A private extension is the shorter identifier entered within the firm's phone system. As a concrete implementation example, Talkroute documents extensions that forward calls to mobile phones, landlines, or other numbers. The identifier and the receiving device therefore answer different questions. ^[9]

A recorded list of names is another distinct choice. It tells callers whom to select without asking them to translate spelling into digits. It is not equivalent to searching a directory. Spoken-name recognition is a separate capability to ask about.

Documentation was checked on October 30, 2026. Open questions require a demonstration. NIST's representative-user, representative-task approach provides the measurement framework. ^[6] The publisher, autoattendant.io, occupies the routing-focused menu side of the comparison. A business that requires searchable names should evaluate a platform documenting that behavior rather than assume every menu supplies it.

Purpose-Based Phone Menus

Capabilities

A short menu asks callers to classify their request: “For billing, press the billing key; for an existing matter, select the client-team option.” The labels should describe an action or destination callers recognize. W3C's voice-menu guidance recommends stating the option before the key used to select it. That is design guidance, not a measured conversion claim. ^[8]

A menu can connect to a person without making that person's name the organizing principle. Vonage's Virtual Receptionist documentation, for example, describes trying another extension before continuing through the automated flow. ^[10] This makes a general-help destination a possible bridge between a structured menu and a conversation.

The forwarding destination is separate from the selection method. Talkroute's extension instructions describe sequential or simultaneous ringing choices. ^[9] A buyer evaluating a business phone menu forwarding to cell phones should test the actual receiving mobiles, not infer delivery behavior from the wording of a directory feature.

Adoption

Choose a purpose menu as the initial candidate when callers can explain their need but do not consistently know a staff member. In a (Hypothetical Example), a small bookkeeping practice might organize calls around new enquiries, existing work, and billing. This is a proposed design for testing, not a claim about how bookkeeping firms generally behave.

Collect a sample of caller tasks before rewriting the greeting. Record what the caller knew at entry, whether the first destination could help, and whether someone had to transfer the call. GOV.UK's usability benchmarking guidance recommends measuring completion time alongside success and abandonment; the same measurement categories can inform a telephone trial without becoming telephone-specific benchmarks. ^[11]

Suggested menu-design checks are:

• Purpose: Each label describes a request the caller can recognize.
• Ownership: Each destination has an accountable receiving person.
• Repeat: Callers can understand how to hear their choices again.
• Silence: The intended no-input route reaches a useful destination.
• Recovery: A caller who cannot choose still has a useful next step.
• Availability: The destination makes sense when the business is closed.

The recovery requirement needs configuration, not merely a spoken promise. Dialpad's department-routing documentation says an administrator can select a default menu option instead of its default no-selection ending behavior. ^[12]

Strengths and Limitations

Purpose routing avoids requiring staff spelling. Its weakness is a mismatch between the menu's categories and the caller's knowledge. A repeat caller who wants a particular adviser may have no reason to choose “new work” or “existing work.” The proposed remedy is a tested named-person route, a published extension, or a selective direct number.

Do not assume familiar callers can skip a greeting everywhere. 3CX expressly documents entering an extension while its digital receptionist prompt plays. ^[13] Dialpad separately documents a recorded-message action whose keypad input is ignored during playback; that limitation applies to that action, not automatically to every Dialpad greeting or directory. ^[12] Demonstrate the specific call path before treating interruption as a requirement met.

Dial-by-Name Directories

Capabilities

A keypad directory translates letters into telephone-keypad digits and identifies matching destinations. Dialpad's current instructions specify the recipient's first name, while Vonage describes entering the first letters of either first or last name. Those are different caller instructions, not interchangeable feature ticks. ^{[3] [14]}

Some directory experiences use recorded identification rather than simply reading a profile. Vonage plays the extension's recorded Name Greeting. ^[14] Grandstream's dedicated directory guide describes spelling a name letter by letter by default, with recorded voicemail greetings available as an alternative. ^[15] Ask what an external caller actually hears, particularly where similar names need an identifying label.

GoTo also documents a choice between recorded names and automated spelling. ^[16] RingCentral allows first- or last-name search configuration. ^[17] These are demonstration questions for the buyer, rather than reasons to assume all directories announce and recognize names identically.

A speech directory adds another input method. Microsoft's documentation says keypad entry remains usable with speech enabled. ^[5] That supports testing both methods; it does not establish recognition accuracy for a particular accent, language, noisy street, or handset.

Adoption

Add a directory candidate when repeat callers identify the person but do not know an extension, and the firm's task sample shows that the existing menu makes them seek assistance. A caller who knows only “billing” still needs a purpose route or a person who can interpret the request.

Before procurement, define the public roster:

• Client-facing staff: Include people callers reasonably expect to request.
• Contractors: Name an owner who confirms availability and removal.
• Internal roles: Decide explicitly whether callers should find them.
• Aliases: Test preferred names and the names clients actually use.
• Single names: Check whether required profile fields permit the intended listing.
• Departures: Assign a replacement route and an external removal test.

These are buyer decisions, not assertions that every product provides each control. Grandstream explicitly lets administrators select extensions for directory membership. ^[15] Vonage's prerequisites include both first and last name fields, so a one-name contact needs an implementation-specific answer. ^[14]

Strengths and Limitations

A directory reduces the caller's need to know the internal number, but it introduces spelling, inclusion, and matching tasks. In 3CX, users without a recorded self-identification message are inaccessible through Route by Name. A profile existing in the system is therefore insufficient evidence of discoverability. ^[13]

Matching behavior also differs. Grandstream's By Order mode lets callers advance to the next matching result with the star key. ^[15] Microsoft can append department information to distinguish people with the same name. ^[5] Neither behavior should be assumed for a vendor whose opened instructions leave collisions unspecified.

Administrative authority and public visibility must be reviewed separately. Microsoft's setup documentation describes authorized users who can change the attendant, distinct from directory-scope configuration. ^[7] Giving an office manager configuration access does not by itself answer which people external callers should discover. RingCentral lets administrators choose included extensions, and GoTo documents inclusion and exclusion choices. ^{[17] [16]}

Feature Comparison

Table 1 maps caller knowledge to the action to test. The entries are proposed decision rules, not measurements of caller preference. They follow the representative-task approach described by NIST. ^[6]

The important distinction is the task, not how sophisticated the technology sounds. A caller may move between rows during a call. A directory should therefore be reachable without making it the only way to get help. W3C recommends simple recovery and human assistance after persistent errors. ^[8]

Table 2 compares documented implementations, checked for this report on its publication date. “Confirm” identifies a procurement question left unanswered by the reviewed pages. The publisher row discloses its position; it is not evaluated as a directory product.

The table compares routing behavior rather than claiming equivalent bundles. In particular, Zoom licenses the included users, while Microsoft's prerequisites distinguish phone and resource-account licensing. ^{[20] [19]} Do not multiply an advertised entry price by staff count until the vendor confirms that the required directory, external calling, and receiving destinations are included.

Other implementations reinforce the need for precise questions. Webex documents direct extension entry without first choosing a menu item. ^[1] Vonage documents a name greeting and a result for each extension where a user has several extensions. ^[14] These details can change the caller's task even when both products advertise a directory.

Performance and Benchmarks

What to measure

The reviewed sources do not establish a controlled performance ranking for small-team menus versus directories. The proposed evaluation measures successful routing, time to a useful destination, wrong destinations, and recovery after an error. NIST supports measuring task time, errors, and successful completion, rather than reducing usability to speed alone. ^[6]

Define the finish event before testing. “The correct handset starts ringing” differs from “the intended person answers.” Record them separately when both matter. Otherwise a slow answer can be mistaken for slow navigation, or a quick transfer to the wrong destination can look successful.

Use the same task wording when comparing designs. GOV.UK recommends keeping tasks and questions consistent between benchmarking rounds. ^[11] The following editable log leaves observations blank deliberately:

task,caller_knowledge,expected_person_or_route,observed_behavior,prompts,elapsed_seconds,result,owner
named_person,name_only,,,,,,
billing,purpose_only,,,,,,
known_extension,extension_only,,,,,,
unknown_contact,neither,,,,,,
no_match,misspelled_name,,,,,,
after_hours,purpose_only,,,,,,
relay_call,agreed_test_task,,,,,,

This is an acceptance-test template, not a dataset. Set a local pass rule before the trial. A useful rule might require the correct route and an intelligible recovery path; the firm should choose any timing target from its own needs and observations.

Run the acceptance test

Test through the external business number, not only an internal preview. Use the receiving mobile phones and realistic caller instructions. Talkroute's documentation makes clear that extensions can forward to existing phone numbers; the test must therefore include that delivery leg. ^[9]

• Known person: Enter the name exactly as a returning caller knows it.
• Collision: Request each person whose keypad prefix matches another.
• Wrong spelling: Enter a plausible misspelling and observe recovery.
• Unknown name: Seek help without choosing a staff member.
• Known extension: Enter it during the greeting and document the outcome.
• No input: Remain silent through the configured menu behavior.
• Unavailable person: Check the intended unanswered-call destination.
• Closed business: Call outside the configured opening hours.

Check the difference between typing letters on a smartphone screen and entering the digits a directory expects. Zoom's caller instructions specify one keypad press for each letter and provide controls to repeat results or start another search. ^[21] Include those actions in the test rather than assuming callers will discover them unaided.

For no input or invalid input, verify the configured destination instead of assuming an operator will answer. 3CX explicitly provides configuration for those cases. ^[22] Microsoft's operator feature requires its operator key assignment; merely filling an operator field is insufficient. ^[7]

Test directory availability separately from after-hours behavior. Microsoft's setup guide permits different greetings and handling outside business hours. ^[7] Decide whether a named-person search should remain reachable then, and whether its destination matches the firm's coverage arrangements.

Access, language, and relay calls

Invite a range of callers to try the tasks, including people who need more time or prefer keypad input. W3C specifically recommends testing voice systems with people with different learning and cognitive disabilities. Its guidance is supplemental design advice, not a claim that this worksheet establishes legal compliance. ^[8]

Plan a relay-call test with an agreed participant and record the actual interaction. In Canada, the Canadian Radio-television and Telecommunications Commission (CRTC) describes message relay as text-assisted calling with a relay operator. ^[23] In the United States, the District of Columbia regulator explains that a communication assistant relays messages between the parties. ^[24] Those descriptions support including relay in the test plan, not a universal timeout specification.

Check multilingual expectations explicitly. Zoom's directory documentation names British and American English support. ^[20] A translated greeting alone therefore cannot establish that name search, pronunciation, and recovery work in every language the firm serves.

Data Analysis and Evidence

Calculate collisions before buying

A name collision occurs here when different listed people produce the same chosen keypad prefix. The ITU Telecommunication Standardization Sector (ITU-T) Recommendation E.161 supplies the letter-to-digit mapping used below: ABC on 2, DEF on 3, GHI on 4, JKL on 5, MNO on 6, PQRS on 7, TUV on 8, and WXYZ on 9. (Source: www.itu.int) These digits are calculated from fictional names; they are not vendor performance results.

Table 3 (Hypothetical Example) uses a deliberately selected fictional roster and compares three-letter and four-letter surname prefixes. The labels are proposed distinguishing descriptions, subject to the chosen platform's actual controls.

Calculated from this constructed roster, all four people share a three-digit prefix with someone else; only the Smith pair shares a four-digit prefix. The Brown/Cross example shows that different letters can map to the same digits. Increasing prefix length resolves that pair here, but does not resolve the Smith pair. It is not evidence that a longer prefix always performs better. Talkroute's guide specifically requires one consistent extension length and recommends a submenu for duplicate patterns. ^[2]

Copy the roster into a spreadsheet, choose the documented search convention, and calculate each prefix. Include preferred names, similar surnames, and the shortest published name. A product's actual matching rules decide whether a collision leads to more input, a selection list, or another route. For example, 3CX waits for additional digits when matching users remain. ^[13]

Build the caller-knowledge worksheet

Do not populate the worksheet with assumed industry percentages. Classify the firm's observed caller tasks and keep the inputs editable. NIST's definition of usability testing calls for representative users performing representative tasks. ^[6]

sample_period,call_or_task_id,knew_purpose,knew_person,knew_extension,initial_route,correct_destination,recovery_needed,notes
,,,,,,,,

The categories can overlap: someone may know both a person and a purpose. Keep those fields separate rather than forcing every caller into a single bucket.

• Named-person share: Named-person requests divided by sampled requests.
• Task success: Correct completed routes divided by attempted tasks.
• Recovery frequency: Tasks needing help divided by attempted tasks.
• Observed effort: Measured minutes per change multiplied by monthly changes.

These are proposed calculations, not externally established benchmarks. Leave the administration inputs blank until measured. When reporting the trial, show the underlying counts, task mix, and sampling period alongside any percentages.

Understand evidence limits

Vendor instructions establish supported mechanisms, not independent usability outcomes. Grandstream's dedicated guide describes a three-letter input flow. ^[15] Zoom documents a two-character minimum. ^[20] Those numbers are product rules, not comparable answer-time measurements. NIST's effectiveness concept concerns accurate and complete achievement of a goal, which is why routing accuracy belongs beside elapsed time. ^[6]

Repeat the exercise after a meaningful routing or roster change. GOV.UK recommends periodic benchmarking and comparison of results. ^[11] Preserve the old task wording, but add any newly introduced failure case.

Implications and Future Directions

The main procurement implication is to buy the smallest routing design that passes the firm's caller tasks, including recovery. A directory requirement should be written as observable behavior: whose names can be found, which input is accepted, what happens with competing matches, and where a failed lookup goes. Avoid accepting “directory included” as the whole specification.

Directory lifecycle work needs explicit ownership. Keep this checklist with the call-flow record:

• Inclusion owner: Record who approves public discoverability.
• Departure date: Record when a person should cease receiving calls.
• Update submitted: Record what changed and who submitted it.
• External retest: Record when a caller checked the published number.
• Removal confirmed: Record what the caller actually heard.
• Replacement route: Confirm who receives callers still seeking that person.

Propagation needs observation. Microsoft's reference allows up to 24 hours for address-book updates. ^[5] The setup page discusses up to 36 hours for a new user's listing. ^[7] Treat the difference as documentation of distinct contexts, not permission to promise immediate removal. No equivalent universal timing should be inferred for the other products.

Speech search should undergo the same task test as keypad search. Whether it helps a specific caller remains an observation to collect.

A human receptionist deserves consideration when callers cannot reliably identify a purpose or person, or when the routing decision itself needs discussion. Automation can still support that arrangement. Vonage documents trying an operator destination before continuing through its automated flow. ^[10] A human fallback is useful only when its staffing and unanswered-call arrangements are real.

For US and Canadian firms, include the actual relay interaction in ongoing testing. CRTC explains that the relay operator communicates by text with one party and voice with the other. ^[23] The DC regulator identifies 711 as the nationwide relay access number in its US guidance. ^[24] Test without assuming an ordinary mobile call covers every access need.

Frequently Asked Questions (FAQs)

Is dial-by-extension the same as dial-by-name?

No. The caller supplies a known internal identifier for extension dialing, but supplies letters or a supported spoken name for directory search. Microsoft permits partial names while requiring the full extension. ^[5] Maintain separate acceptance tasks for these inputs.

Can a small business keep a simple menu and add a directory?

A platform may offer the directory as an auto-attendant destination. Webex documents name and extension actions separately. ^[1] Keep the purpose route for callers who do not know staff names, and demonstrate the exact combined flow before purchase.

Does a directory automatically include every employee?

Do not assume that. Zoom requires a Zoom Phone license for included users, while 3CX requires a self-identification message for Route by Name access. ^{[20] [13]} Check intended membership externally after configuration.

Can calls reach existing cell phones?

Some documented implementations forward extensions to mobile numbers. Talkroute explicitly describes that destination type. ^[9] Confirm the menu-to-mobile path, unanswered-call behavior, and closed-hours treatment for the chosen plan rather than treating forwarding as a property of all directories.

Conclusion

The decision between a dial-by-name directory and a phone menu begins with what callers know. Purpose routing is a sensible candidate when callers can name the work they need done. A directory is a sensible candidate when they know the person but lack an extension. Published extensions and selective direct numbers can serve familiar callers, while a human receptionist can interpret requests that resist either classification.

The practical next step is to run the same caller tasks through the candidate designs. Calculate the actual roster's keypad collisions, test each intended listing, try a missing name, and document the recovery route. Keep handset ringing separate from the person answering, and treat a fast wrong destination as a failed task.

The implementation table should guide questions rather than substitute for a demonstration. Search scope, prerequisites, recorded names, interruption, and removal behavior need product-specific answers. Any unresolved requirement belongs in the buying checklist.

Keep the short menu when it works for the firm's observed callers. Add directory functionality when the named-person task merits it and the selected platform passes the acceptance test. Retain direct access selectively, and assign a person to maintain public listings and fallbacks. That produces a decision based on caller needs and measured local behavior, with no assumption that a larger directory is inherently a better phone experience.