How Do You Choose a Marketing Agency?

By Ryan Goering, founder and CEO of BaaDigi. Last updated: September 2026.

Editorial correction: This article removes unsupported agency price ranges, fixed trade sales cycles, invented example results and claims that one provider type always wins. BaaDigi sells the services discussed here; use the same checks on our offer.

Does local proximity matter?

It matters when the work benefits from an in-person activity, such as capturing approved project material or working directly with an office team. Ask whether that activity is included. A nearby address does not establish that the people delivering the work will visit your business.

A remote team may work well when responsibilities, communication and access are clear. Compare actual work samples and the proposed process. A contractor specialist should understand service fit and intake; a generalist may also demonstrate that competence. Do not substitute a label for evidence.

What should you ask during discovery?

1. What problem have you observed in our current marketing?
2. Which records support that diagnosis, and what is missing?
3. Which services and locations would the plan prioritize?
4. What must our office do after a new inquiry arrives?
5. Who approves prices, availability and other public claims?
6. Which tasks are included and which need a separate scope?
7. What will the next report help us decide?

Ask for plain explanations. A useful answer might identify a broken handoff or an incomplete service page. A list of channel names is not yet a strategy. The marketing problem diagnosis helps separate acquisition from handling and measurement.

How should you compare claimed results?

Ask for dates, definitions and the source of each number. Distinguish visits, inquiries, qualified leads and acquired customers. Review the starting conditions and what else changed during the reported period. A real result for another business shows experience with that engagement; it does not guarantee your outcome.

Request references or work examples the agency is permitted to share. Respect client confidentiality rather than assuming every provider can disclose private account records. Where a claim cannot be verified, leave it unverified instead of filling the gap with a persuasive example.

Ask for medians, not just totals. Across 39 client Google Business Profiles BaaDigi tracks, 12 months produced 10,233 calls from the profile (Sep 2025 to Aug 2026), which sounds like a lot per client; the median profile actually got 1.3 calls and 50 impressions a month, because a few large profiles carry the total. An agency quoting only the big number is not lying, but it is not telling you what the typical client got either.

For search work, Google’s hiring guidance recommends examining methods, prior work and measurement, and warns against guaranteed first-place rankings. Use the SEO-specific hiring checklist when that is the proposed scope.

What should the audit before the proposal show?

Any agency worth hiring will audit before it quotes, and you should know what a typical result looks like so a scary report does not stampede you. Across 1,047 free BaaDigi audits completed between March and September 2026 by contractors who chose to run one, the average scores out of 100 were 54.1 overall, 29.0 for SEO, 74.2 for Google Business Profile, 66.8 for page speed and 72.9 for on-page. 93.7% of sites had HTTPS, 82.0% were mobile responsive, 72.9% had an XML sitemap and 59.5% had schema markup. Median mobile PageSpeed was 61, and 22.1% of sites scored under 50 on mobile. The audit found a Google Business Profile for 57.9% of businesses, with a median of 35 reviews and a 4.90 rating. Of 254 audits that ran the check, Perplexity named the business for its own service-and-city query in 59 cases (23.2%), with a median of 3 competitors named ahead of it. These are self-selected prospects, not a random sample, but they show what "your site needs work" usually means: weak SEO, average speed, and near-zero AI visibility. Ask the agency which of those lines it would move first and how it would prove it moved. Our guide to reading an SEO audit covers what each score does and does not mean.

Which responsibilities should be written down?

A longer agreement is not proof of poor service, and month-to-month billing is not proof of competence. Read deliverables, termination terms and dependencies together. Establish how missed work and unresolved problems are handled before there is a dispute.

How do you compare the full price?

Separate setup, recurring agency work, advertising spend, production, tools and usage charges. Identify optional items and the approval process for additional work. Compare the same included work rather than sorting proposals by the monthly headline.

BaaDigi’s Stability Engine is $497 per month plus $995 setup, managing up to $1,600 per month of ad spend that you pay to Google; it is one defined program, not an industry benchmark. Use the marketing budget guide to relate any agency's full cost to your capacity and acquisition records.

What should the first work period produce?

Require a scoped plan with access needs, a baseline, approved priorities and a named owner for each task. Dates should account for dependencies such as missing business facts or website access. The agency should distinguish work it can complete from customer outcomes it cannot promise by a deadline.

Verify important contact routes early. An ad click or website submission is not yet proof that the office received a usable inquiry. Google’s conversion-action documentation illustrates why the configured action matters. Agree on definitions before accepting a lead count.

How can you judge the relationship fairly?

Review completed work, open issues, data completeness and qualified outcomes using consistent definitions. A relevant problem can justify a focused change; a short fluctuation alone may not. Keep seasonality, staffing, campaign changes and unresolved estimates visible when assessing performance.

BaaDigi uses the Predictable Work Dashboard as a transparency tool. Ask which records support each measure and where attribution is incomplete. The marketing scorecard guide explains how to keep observations separate from causal claims.