What a good discovery process looks like

The tell is simple. If a proposal arrives before anyone asked which of your services carries the best margin, you were sold a package rather than given a diagnosis.

That is not always wrong. Packages exist because many businesses need the same first steps. But you should know which one you are buying, because a package priced as a diagnosis is a bad trade.

What is discovery supposed to produce?

A document, not a feeling.

It should contain findings, meaning what was examined and what was found. Prioritized recommendations, with the reason each one is placed where it is. A baseline of current numbers. And a risk list, because the things that could go badly wrong deserve their own section.

The risk list is the part most often missing and the part with the highest value. Listing problems belong there, since a profile suspension takes real work to reverse and knowing you are exposed is worth more than a month of content.

The baseline has one rule attached. It should be your own trailing numbers, not a comparison against a published industry average, because there is no current published average for local business profile performance that survives checking. Compare your ninety days against your prior ninety days and against your named local competitors. Any discovery that opens with "you are below the industry benchmark" is quoting something that either has no disclosed sample or is built on metrics the platform has since retired.

The prioritization is where you learn whether they are any good. Anyone can produce a list of forty issues from an audit tool. Ordering them by what will change your revenue, and saying which twenty-five do not matter for you, requires judgment.

Ask what they decided not to recommend and why. A discovery document with no exclusions is an inventory, not an analysis.

What should they ask that has nothing to do with marketing?

The questions that determine whether any of this will make you money.

Which services or products carry your best margin. Which jobs you want more of and which you take reluctantly. What your capacity looks like, and what happens if calls increase by a third. Where customers come from today, by rough proportion. What you have already tried and what happened.

Then the operational ones. Who answers the phone and when. How quickly you respond to a form submission. What happens to an inquiry that arrives at 7pm on a Saturday.

Those last questions separate an agency planning to make you money from one planning to make you traffic. If the leads already arriving are not being handled, more of them changes nothing, and a good discovery says so before you sign anything.

They should also ask about your time. How much you have, who approves work, and how fast. That belongs in discovery because it determines what plan is realistic, and it overlaps with the questions you should be asking them in the same meeting.

What does the funnel question look like in numbers?

This is the arithmetic a good discovery forces you to do, and you can run it yourself this week.

Start with a fully disclosed example. Patient Prism scored 11,552,668 patient calls across 8,280 dental locations for calendar year 2025. Per 100 calls: 69 reached the front desk and 31 hung up before reaching an agent. Of the calls that connected, 34 were real booking opportunities, 21 booked, and 13 walked away without booking. Of those 13, only 4.8 received a followup call, and 0.7 came back and booked.

Read the last two numbers again. Fewer than one patient in a hundred who reached out and then walked away ever came back through the door. The largest recoverable pool in that funnel is almost entirely unworked, and none of it is a traffic problem.

Now run yours. Pull last month's inbound calls from your phone system. Count how many were answered. Count how many became a booking, a job, or a quote. If you cannot produce either number, that is the first finding of your discovery, and it did not require an agency to reach.

Then the number that decides the engagement. ServiceTitan's platform data puts the typical home services shop at a 42% call booking rate, with shops under five technicians at 24% and shops with 25 or more at 59%. The same analysis found that every 5% improvement in booking rate is worth roughly $100,000 in additional revenue for a shop with five to fourteen technicians, achievable with less than one additional booked call per weekday.

So if you answer 70 calls a month and book 30 of them, the question is not how to get 100 calls. It is what happened to the other 40.

Two caveats, and they matter. Patient Prism's cohort is made up of practices that bought call intelligence software, so the true market abandonment rate is plausibly worse rather than better. ServiceTitan's gradient by shop size says something uncomfortable and true: missed calls are usually a capacity problem, not a discipline problem. The small operator is on a roof or under a sink. A discovery that treats it as laziness has diagnosed the wrong thing.

Should I pay for discovery?

Often yes, and the reason is structural rather than moral.

Free discovery is a sales call. Both parties know it, and the incentive is to arrive at a recommendation that matches what the agency sells. That is not dishonesty, it is gravity.

Paid discovery changes the incentive and changes what you get. You should own the document. You should be able to take it to another vendor, or execute it yourself, with no obligation to continue. Ask for that in writing before you pay, and if the answer is no, it was not really discovery.

A fair paid discovery has a defined scope, a fixed price, and a delivery date. If it converts into an engagement, having the fee credited is a reasonable thing to negotiate, though it slightly reintroduces the incentive problem.

Which discovery findings should you push back on?

The urgent technical requirement you had never heard of. Some of those are real. Some are products.

Structured data is the most common example. Google's own guidance states that structured data is not required for generative AI features and that there is no special schema markup you need to add, and it names artificial content chunking, rewriting content for AI systems, machine-readable AI text files, and seeking inauthentic mentions as things that will neither harm nor help visibility. So a discovery that leads with schema as an AI necessity is selling a product, and Google states plainly that it is not required.

Two related items to check in the same paragraph. FAQ markup has been restricted to well-known authoritative government and health websites since August 2023, and How-to rich results were retired on desktop in September of the same year, so neither produces a rich result for an ordinary local business. And a plugin that scrapes your Google reviews and marks them up as an aggregate rating on your own homepage violates Google's structured data guidelines, which prohibit marking up reviews about your own business on your own site.

None of this means structured data is worthless. Local business and organization markup does real work for entity disambiguation and for genuine rich results, and Microsoft has said publicly that schema helps its models understand web content. The honest version is that it is infrastructure, not an AI visibility play, and a discovery document that says so is a better sign than one promising otherwise.

How long should it take?

For a single-location business, days rather than weeks. Two to three hours of your time, a few days of theirs, a document at the end.

For multiple locations or a complex service mix, longer, and it should include interviews with people other than you. Your front counter staff know which questions customers ask. Your sales lead knows which inquiries close. An agency that only talks to the owner is missing most of the useful information.

This is one area where a larger agency genuinely does more, and it is worth saying plainly. A dedicated researcher running a week of structured interviews across your team, plus a specialist reviewing the ad account and another reviewing the site build, will surface things a two-person shop working from one conversation will not. If you have the complexity to justify that, buy it. A small shop that tells you otherwise is defending its own shape.

What should not happen is discovery that runs for months with no artifact. That is billable delay. Ask for the delivery date in writing, and keep the commercial terms separate from the discovery scope so you are not committed by momentum, which is one more reason to settle the contract terms deliberately rather than at the end of a long process.

What to do this week

Ask any candidate for their discovery agenda before the meeting. What they intend to ask, and what you will receive at the end.

Read the agenda for business questions. Count how many are about margin, capacity, and how inquiries are handled today. If the answer is none, you will get a marketing plan with no idea whether marketing is your constraint.

Pull your own call and inquiry counts for last month and run the funnel arithmetic above. Walk into the meeting with those two numbers already written down. It changes the conversation immediately, and it tells you within ten minutes whether the person across the table wants to talk about them.

Then ask a reference about it directly. "What did their discovery find that you did not already know?" That is one of the most useful questions available, and reference calls give real answers when you ask about process rather than results.

Questions about a discovery process you are about to start?

Email me at eric@seod.com with the agenda you were sent for your kickoff or discovery call, and I will tell you what is missing from it and the three business questions nobody appears to be planning to ask. You take that into your own meeting. I am not looking to run it and there is no message afterward.

There is more on hiring and working with agencies if you are earlier than that.