Routing leads from four channels without dropping any

This is unglamorous work. It is also the highest return hour available to most local businesses, because the failure it fixes is not slow response, it is no response.

The Harvard Business Review audit of 2,241 US companies found that 23% never responded to a web generated lead at all. Not late. Never. One disclosure travels with that number and almost nobody gives it: co-author David Elkington founded InsideSales.com, whose platform supplied the underlying data, and HBR is a magazine rather than a peer reviewed journal. It is high quality vendor research published under an editorial masthead. The finding still holds up, and the 23% is the part of it worth acting on.

Which channels are actually in play?

More than four, usually, which is why the table matters.

The phone, obviously. Then web forms, which are often more than one form on more than one page, each going to a different address depending on who built them and when.

Google Business Profile messages, the one I find orphaned most often. It arrives in an app on a phone belonging to whoever set up the profile, sometimes a former employee.

Texts to your main line, now a real channel whether or not you decided it would be.

Then whatever else exists in your category: chat widgets, social messages, marketplace platforms, lead aggregators.

List every one of them before you improve any of them. You cannot route what you have not named.

Which channels actually produce revenue?

Weight the table by where your business really comes from, because the answer is more concentrated than most owners expect.

The best evidence in any local vertical comes from dental. Patient Prism attributed 5,022,887 patient calls across 8,280 locations and found that 90% of attributable calls came from the Google ecosystem. The Google Business Profile alone accounted for 54%, organic search and Google Ads together another 36%, and Meta plus print combined came to under 3%.

That is dental, and the cohort is practices sophisticated enough to buy call intelligence software, so say so when you use it. The shape is still worth taking seriously. If your routing table gives equal weight to Instagram messages and Business Profile calls, it is weighted for the wrong century.

Two things complicate that picture. Sterling Sky's field data, drawn from Jepto's compilation of 179 Google Business Profiles across 34 US law firms over two years, shows clicks to call declining even where rankings held steady, with the loss concentrated on mobile where the call icon lives. And the AI powered local packs now appearing on roughly 7% of tracked local keywords show one or two businesses instead of three and carry no call buttons at all.

The phone is getting a smaller share of a shrinking set of buttons. That is not an argument for ignoring it. It is an argument for making sure every other channel has an owner, because the traffic that used to become a call is becoming something else.

Which one has no owner?

Test each channel yourself, as a stranger, and time the response.

Submit each web form. Send a Google Business Profile message. Text your main line. Use the chat widget. Message the Instagram account. Call the number on your truck. Do it on a Saturday and again on a Tuesday afternoon.

The one that never comes back is your project. It is rarely the phone, because the phone has a person near it and it makes a noise. It is usually a form going to an address nobody monitors, or a profile message landing in an app on a phone that is on silent.

Once you find it, do not build a process. Change the destination so it lands where somebody already looks daily, then assign a name to it.

How do you turn the test into a number?

What does the routing table look like?

Four columns and one row per channel. It fits on a single page and it should live where the team can see it.

The columns that get skipped are owner and response window, and they are the two that make it real. "The team" is not an owner. A response window without a name attached is a wish.

Consolidate destinations wherever you can. Four channels feeding one shared inbox beats four channels feeding four places, even when each is technically monitored. Then handle duplicates deliberately. Someone who fills in a form and calls twenty minutes later is one lead, and treating them as two is how a caller gets asked the same questions twice.

One more design note that changes the response window column. A TNS survey in July 2022 found that 75% of Americans never answer calls from a number they do not recognize. Your callback is an unrecognized number. Where a channel lets you resolve the request in the channel it arrived on, do that instead of promising a call back. It is a better experience and it converts better.

The phone row deserves the most thought because it has the most options. The tradeoffs between AI answering, voicemail, and a live answering service determine what goes in the destination column, and that decision should follow the volume, not the other way around.

Do the rules travel across channels?

No, and this is the part that gets businesses into trouble.

Each channel carries its own obligations and they do not transfer. What a phone agent may say is not what an autoresponder may say. A restaurant learns this fast, because the hard line on allergen questions applies to a chat widget exactly as it applies to a voice agent, and chat widgets are usually built by someone who never heard the rule.

Three specifics worth putting in front of your own attorney rather than working out from a blog post.

Recording. If your phone destination records or transcribes, California requires the consent of all parties before recording begins, and several other states have all party rules. That notice belongs in the first line of the greeting.

Texting. If a row sends an automated text, US carriers require business messaging traffic to be registered before it delivers reliably, and federal rules let a consumer revoke consent by any reasonable method. Suppression has to be global across your whole table. A caller who says "stop texting me" out loud to your phone agent must not keep receiving messages from your form autoresponder, and in most deployments those are two systems that never speak to each other.

Automated outbound. If any row proposes an AI voice ringing people back rather than answering them, that is outbound artificial voice calling, which the FCC has held requires the prior express consent of the called party. It is a different legal object from answering the phone, and it deserves its own line in the conversation with counsel.

None of that is legal advice. It is the list of questions that makes a thirty minute meeting productive.

What breaks a routing table?

Three things, and all of them are predictable.

Staff changes. The owner column goes stale the moment somebody leaves. Reassign it the same week, and check the profile messages app, because it is usually tied to a personal login.

New marketing. Every new campaign, page, or listing creates a new arrival path. If the table is not updated when the campaign launches, the campaign has an unowned channel from day one.

Volume shifts by category. CallRail's analysis of 1.1 million de-identified conversations put missed call rates at 32% in health care, 28% in legal, 14% in home services, and 9% in real estate. If your phone row is under strain, the table should reflect it before the strain shows up in revenue.

Review the table quarterly. It takes fifteen minutes and it catches all three.

The phone row has its own warning signs, and the signs that a business is losing customers on the phone show up in the outcome column before they show up in revenue.

Also count offline. If you run vehicle graphics or yard signs, those calls belong in the table too, and whether offline advertising is doing anything measurable is a question you can only answer if the calls are attributed when they arrive.

What to do this week

Write the list of every channel. Include the ones you suspect nobody is watching.

Test each one as a stranger, on a Saturday and a Tuesday, and time the responses.

Run the six step count above and produce a response rate per channel. The lowest number is your project.

Build the four column table. Put a person's name in every owner cell.

Change the destination on whichever channel failed, so it lands somewhere already monitored.

Then set a monthly count of arrivals and outcomes per channel. Do it honestly, because measuring conversion honestly at small traffic volumes is harder than it looks and the temptation to flatter yourself is strong.

Questions about where your leads are going?

Email me at eric@seod.com with a list of every place a customer can reach you, including the ones you are unsure about. I will tell you which channel most likely has no owner, what to test first, and the destination change I would make.

I answer these personally. In most businesses I can guess the orphaned channel before I finish reading the list, and I am usually right.

More on the rest of the chain sits in the AI phone and lead response library.